Upload project files excluding raw data folder

.gitattributes

@@ -58,3 +58,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 # Video files - compressed
 *.mp4 filter=lfs diff=lfs merge=lfs -text
 *.webm filter=lfs diff=lfs merge=lfs -text
+paper/.ipynb_checkpoints/main-checkpoint.pdf filter=lfs diff=lfs merge=lfs -text
+paper/main.pdf filter=lfs diff=lfs merge=lfs -text

.ipynb_checkpoints/PLAN-checkpoint.md

@@ -0,0 +1,1175 @@
+Dưới đây là **kế hoạch hoàn chỉnh phiên bản mạnh hơn** cho paper phát hiện té ngã, có dataset link, mô tả dataset, method lấy từ các paper khác, proposed method, thí nghiệm, bảng kết quả và lộ trình chạy.
+
+# 1. Hướng paper chốt
+
+## Tên paper đề xuất
+
+**DynaFall: Robust Skeleton-Based Fall Detection via Keypoint Dynamics and Graph-Based Benchmarking**
+
+Tên này tốt hơn tên cũ vì nó thể hiện 2 điểm:
+
+1. Có **model riêng**: DynaFall.
+2. Có **benchmark với các method mạnh từ paper khác**: ST-GCN, 2s-AGCN, CTR-GCN, PoseC3D, TCNTE-style.
+
+Ý tưởng chính:
+
+```text
+RGB Video
+→ YOLO Pose
+→ 17-keypoint skeleton sequence
+→ Joint / Bone / Dynamics features
+→ Compare with strong skeleton methods
+→ Proposed DynaFall-GCN
+→ Fall / Non-fall
+```
+
+Paper sẽ không còn đơn giản là “dùng CNN/LSTM/Transformer để phân loại té ngã”, mà thành:
+
+> Một framework phát hiện té ngã dựa trên skeleton, chạy lại các method skeleton action recognition mạnh trên cùng dataset, sau đó đề xuất model chuyên biệt cho fall detection bằng keypoint dynamics và confidence-aware dropout.
+
+---
+
+# 2. Dataset dùng trong paper
+
+Tôi khuyên dùng **2 dataset chính + 1 dataset backup/extension**.
+
+## Dataset chính 1: UR Fall Detection Dataset — URFD
+
+**Link dataset:** UR Fall Detection Dataset. ([Fenix][1])
+
+**Mô tả:** URFD gồm **70 sequences**, trong đó có **30 fall sequences** và **40 activities of daily living — ADL sequences**. Fall events được ghi bằng **2 Microsoft Kinect cameras** kèm dữ liệu accelerometer; ADL events được ghi bằng camera 0 và accelerometer. ([Fenix][1])
+
+**Vì sao nên dùng:**
+
+* Nhỏ, dễ chạy nhanh.
+* Rất phổ biến trong fall detection.
+* Phù hợp để debug pipeline trước.
+* Có RGB video để trích xuất YOLO pose.
+
+**Vai trò trong paper:**
+
+```text
+Dataset kiểm tra nhanh + benchmark chính.
+```
+
+---
+
+## Dataset chính 2: Multiple Cameras Fall Dataset — MCFD
+
+**Link dataset:** Multiple Cameras Fall Dataset. ([Iro Montéal][2])
+
+**Mô tả:** MCFD có **24 scenarios** được ghi bằng **8 IP video cameras**. **22 scenarios đầu** chứa fall và các sự kiện gây nhầm lẫn; **2 scenarios cuối** chỉ chứa các sự kiện gây nhầm lẫn. ([Iro Montéal][2])
+
+**Vì sao nên dùng:**
+
+* Có nhiều góc camera.
+* Phù hợp để kiểm tra robustness.
+* Paper nhìn mạnh hơn URFD-only.
+* Có các tình huống dễ gây nhầm: ngồi, cúi, nằm, chuyển động gần giống fall.
+
+**Vai trò trong paper:**
+
+```text
+Dataset kiểm tra multi-view + robustness.
+```
+
+---
+
+## Dataset backup/extension: UP-Fall Detection Dataset
+
+**Link dataset:** UP-Fall / Challenge UP data page. ([Google Sites][3])
+
+**Mô tả:** Trang Challenge UP mô tả UP-Fall là dataset lớn cho fall detection, gồm **11 activities**, **3 trials per activity**, thực hiện bởi **12 subjects**, bao gồm 6 hoạt động hằng ngày và 5 kiểu té ngã, thu bằng wearable sensors, ambient sensors và vision devices. ([Google Sites][3]) Paper gốc về UP-Fall cũng nhấn mạnh đây là dataset multimodal cho fall detection và human activity recognition. ([MDPI][4])
+
+**Khi nào dùng:**
+
+* Nếu MCFD tải khó hoặc xử lý annotation khó.
+* Nếu muốn paper mạnh hơn và có thêm dataset thứ 3.
+* Nếu muốn hướng “multimodal dataset nhưng chỉ dùng RGB-derived pose”.
+
+**Khuyến nghị thực tế:**
+Paper 8 trang cho hội nghị Q4 thì dùng **URFD + MCFD** là đủ. UP-Fall để backup.
+
+---
+
+# 3. Dataset protocol chốt
+
+Trong paper ghi rõ:
+
+```text
+Although some datasets provide depth, accelerometer, or multimodal signals, this study only uses RGB videos to extract human skeleton keypoints. This ensures a consistent vision-based evaluation protocol across all datasets.
+```
+
+Tức là:
+
+| Dataset          | Dữ liệu gốc có gì         | Mình dùng gì               |
+| ---------------- | ------------------------- | -------------------------- |
+| URFD             | RGB, depth, accelerometer | Chỉ RGB → YOLO Pose        |
+| MCFD             | Multi-camera RGB video    | RGB → YOLO Pose            |
+| UP-Fall nếu dùng | wearable, ambient, vision | Chỉ RGB/vision → YOLO Pose |
+
+Điểm này rất quan trọng để reviewer thấy so sánh công bằng.
+
+---
+
+# 4. Proposed method
+
+## Tên model
+
+**DynaFall-GCN**
+
+## Ý tưởng
+
+Model của mình không chỉ dùng skeleton thô, mà có 3 stream:
+
+```text
+Stream 1: Joint stream
+- x, y, confidence của 17 keypoints
+
+Stream 2: Bone stream
+- vector xương giữa các keypoints
+
+Stream 3: Dynamics stream
+- velocity
+- acceleration
+- torso angle
+- hip drop
+- body aspect ratio
+- center-of-body motion
+```
+
+Sau đó dùng:
+
+```text
+Graph-temporal encoder
++ confidence-aware keypoint dropout
++ classification head
+```
+
+---
+
+# 5. Pipeline tổng thể
+
+```text
+Input RGB Video
+    ↓
+Frame Sampling
+T = 32 frames
+    ↓
+YOLOv8/YOLO11 Pose Extraction
+17 COCO keypoints per frame
+    ↓
+Pose Normalization
+bbox-centered normalization
+    ↓
+Feature Construction
+joint + bone + dynamics
+    ↓
+Comparative Methods
+ST-GCN / 2s-AGCN / CTR-GCN / PoseC3D / TCNTE-style
+    ↓
+Proposed DynaFall-GCN
+joint stream + bone stream + dynamics stream
+    ↓
+Robustness Evaluation
+missing keypoints / lower-body occlusion / low-confidence mask
+    ↓
+Fall / Non-fall Prediction
+```
+
+---
+
+# 6. Các method từ paper khác để chạy lại
+
+Đây là phần giúp paper “hay hơn” và không bị đơn giản.
+
+## 6.1 LSTM baseline
+
+Dùng để kiểm tra sequence baseline.
+
+Input:
+
+```text
+32 frames × 17 keypoints × 3
+→ flatten
+→ 32 × 51
+```
+
+Vai trò:
+
+```text
+Basic temporal baseline.
+```
+
+---
+
+## 6.2 ST-GCN
+
+**Paper:** Spatial Temporal Graph Convolutional Networks for Skeleton-Based Action Recognition, AAAI 2018.
+
+ST-GCN mô hình hóa skeleton như một graph không-thời gian, trong đó keypoints là nodes, bones là edges, và mô hình học cả quan hệ không gian giữa khớp và quan hệ thời gian giữa frames. ([AAAI Publications][5])
+
+Áp dụng vào paper:
+
+```text
+Input: N × C × T × V × M
+C = 3
+T = 32
+V = 17
+M = 1
+Output = Fall / Non-fall
+```
+
+Vai trò:
+
+```text
+Classic graph-based skeleton baseline.
+```
+
+---
+
+## 6.3 2s-AGCN
+
+**Paper:** Two-Stream Adaptive Graph Convolutional Networks for Skeleton-Based Action Recognition, CVPR 2019.
+
+2s-AGCN mạnh hơn ST-GCN vì graph topology có thể được học thích nghi, không cố định hoàn toàn. Method này cũng dùng hai stream: **joint stream** và **bone stream**, trong đó bone stream chứa thông tin độ dài và hướng xương. ([arXiv][6])
+
+Áp dụng:
+
+```text
+Joint stream: keypoint coordinates
+Bone stream: bone vectors
+Final prediction: score fusion
+```
+
+Vai trò:
+
+```text
+Adaptive graph + bone information baseline.
+```
+
+---
+
+## 6.4 CTR-GCN
+
+**Paper:** Channel-wise Topology Refinement Graph Convolution for Skeleton-Based Action Recognition, ICCV 2021.
+
+CTR-GCN học graph topology theo từng channel, giúp mô hình học các quan hệ khớp khác nhau cho từng loại feature. Đây là một baseline GCN mạnh cho skeleton action recognition. ([arXiv][7])
+
+Áp dụng:
+
+```text
+Input: joint + bone skeleton sequence
+Graph: COCO 17-keypoint topology
+Output: Fall / Non-fall
+```
+
+Vai trò:
+
+```text
+Strong graph-based baseline.
+```
+
+---
+
+## 6.5 PoseC3D / PoseConv3D
+
+**Paper:** Revisiting Skeleton-based Action Recognition, CVPR 2022.
+
+PoseC3D/PoseConv3D không dùng graph sequence trực tiếp, mà chuyển pose thành **3D heatmap volume**. Paper gốc cho rằng cách này hiệu quả hơn trong học spatiotemporal features, robust hơn với pose estimation noise và generalize tốt hơn trong cross-dataset settings. ([CVF Open Access][8])
+
+Áp dụng:
+
+```text
+YOLO keypoints
+→ 3D pose heatmap volume
+→ 3D CNN
+→ Fall / Non-fall
+```
+
+Vai trò:
+
+```text
+Heatmap-based skeleton baseline.
+```
+
+---
+
+## 6.6 TCNTE-style
+
+**Paper:** Real-time skeleton-based fall detection algorithm based on Temporal Convolutional Networks and Transformer Encoder, 2025.
+
+TCNTE kết hợp **Temporal Convolutional Network** với **Transformer Encoder** cho fall detection thời gian thực, dùng YOLOv8 pose extraction, BoT-SORT tracking, và weighted focal loss để xử lý mất cân bằng lớp. ([ScienceDirect][9])
+
+Áp dụng:
+
+```text
+TCN + Transformer Encoder
+Input: skeleton sequence
+Loss: weighted focal loss
+Output: Fall / Non-fall
+```
+
+Vai trò:
+
+```text
+Recent fall-specific temporal baseline.
+```
+
+---
+
+# 7. Bảng method chính trong paper
+
+Nên đưa bảng này vào phần **Experimental Setup**.
+
+| Method           | Loại              | Nguồn gốc           | Input                   | Vai trò                     |
+| ---------------- | ----------------- | ------------------- | ----------------------- | --------------------------- |
+| LSTM             | Sequence model    | baseline phổ biến   | joint sequence          | baseline cơ bản             |
+| ST-GCN           | Graph-based       | AAAI 2018           | joint graph             | skeleton baseline kinh điển |
+| 2s-AGCN          | Adaptive GCN      | CVPR 2019           | joint + bone            | baseline mạnh hơn ST-GCN    |
+| CTR-GCN          | Refined GCN       | ICCV 2021           | joint + bone            | strong GCN baseline         |
+| PoseC3D          | Heatmap 3D CNN    | CVPR 2022           | pose heatmap            | khác hướng GCN              |
+| TCNTE-style      | TCN + Transformer | fall detection 2025 | skeleton sequence       | fall-specific baseline      |
+| **DynaFall-GCN** | Proposed          | ours                | joint + bone + dynamics | model chính                 |
+
+---
+
+# 8. Điểm mới của DynaFall-GCN so với các method kia
+
+| Thành phần                        | ST-GCN         | 2s-AGCN        | CTR-GCN        | PoseC3D         | TCNTE                     | DynaFall-GCN |
+| --------------------------------- | -------------- | -------------- | -------------- | --------------- | ------------------------- | ------------ |
+| Joint stream                      | Có             | Có             | Có             | Heatmap         | Có                        | Có           |
+| Bone stream                       | Không          | Có             | Có             | Không trực tiếp | Không rõ                  | Có           |
+| Velocity/Acceleration             | Không          | Không          | Không          | Không trực tiếp | Có thể có motion temporal | Có rõ        |
+| Torso angle                       | Không          | Không          | Không          | Không           | Không                     | Có           |
+| Hip drop                          | Không          | Không          | Không          | Không           | Không                     | Có           |
+| Body aspect ratio                 | Không          | Không          | Không          | Không           | Không                     | Có           |
+| Confidence-aware dropout          | Không          | Không          | Không          | Không           | Không rõ                  | Có           |
+| Robustness test missing keypoints | Không bắt buộc | Không bắt buộc | Không bắt buộc | Có thể robust   | Không bắt buộc            | Có           |
+
+Thông điệp chính:
+
+> Các method skeleton action recognition mạnh học action nói chung, còn DynaFall-GCN đưa thêm fall-specific dynamics và robustness training để phù hợp hơn với bài toán té ngã.
+
+---
+
+# 9. Chi tiết xử lý dữ liệu
+
+## Step 1: Download dataset
+
+Tải:
+
+1. **URFD** từ trang chính thức. ([Fenix][1])
+2. **MCFD** từ trang dataset của Université de Montréal. ([Iro Montéal][2])
+3. **UP-Fall** chỉ dùng nếu cần backup hoặc mở rộng. ([Google Sites][3])
+
+---
+
+## Step 2: Chuẩn hóa thư mục
+
+```text
+fall_project/
+├── data/
+│   ├── raw/
+│   │   ├── URFD/
+│   │   ├── MCFD/
+│   │   └── UPFall_optional/
+│   ├── poses/
+│   ├── clips/
+│   └── splits/
+├── src/
+├── configs/
+├── results/
+└── paper/
+```
+
+---
+
+## Step 3: Trích xuất pose
+
+Dùng:
+
+```text
+YOLOv8n-pose hoặc YOLO11n-pose
+```
+
+Output mỗi frame:
+
+```text
+17 keypoints × 3
+x, y, confidence
+```
+
+Nếu có nhiều người trong frame:
+
+```text
+Chọn người có bbox lớn nhất.
+```
+
+Lý do:
+
+```text
+Fall datasets thường chỉ có một subject chính.
+```
+
+---
+
+## Step 4: Normalize keypoints
+
+Với mỗi frame:
+
+```text
+x_norm = (x - bbox_center_x) / bbox_width
+y_norm = (y - bbox_center_y) / bbox_height
+confidence giữ nguyên
+```
+
+Output:
+
+```text
+T × 17 × 3
+```
+
+Trong đó:
+
+```text
+T = 32 frames
+```
+
+---
+
+## Step 5: Tạo clip
+
+Dùng:
+
+```text
+clip length = 32 frames
+stride = 16
+```
+
+Nếu video ngắn:
+
+```text
+uniform sampling hoặc padding frame cuối
+```
+
+Quan trọng:
+
+```text
+Chia train/val/test theo video trước, rồi mới tạo clip.
+Không random clip trực tiếp vì sẽ leak dữ liệu.
+```
+
+---
+
+# 10. Feature engineering cho proposed model
+
+## 10.1 Joint features
+
+```text
+x, y, confidence
+```
+
+Shape:
+
+```text
+32 × 17 × 3
+```
+
+---
+
+## 10.2 Bone features
+
+Dựa trên COCO skeleton edges:
+
+```text
+bone_vector = child_joint - parent_joint
+```
+
+Các cạnh chính:
+
+```text
+nose-eye-ear
+shoulder-elbow-wrist
+shoulder-hip
+hip-knee-ankle
+left-right shoulder
+left-right hip
+```
+
+Shape:
+
+```text
+32 × 17 × 3
+```
+
+---
+
+## 10.3 Dynamics features
+
+Tạo thêm:
+
+```text
+velocity = p_t - p_{t-1}
+acceleration = v_t - v_{t-1}
+torso angle
+hip drop
+body aspect ratio
+center-of-body velocity
+```
+
+Fall thường có:
+
+```text
+torso angle thay đổi nhanh
+hip y-coordinate tụt nhanh
+bbox height/width thay đổi
+center-of-body velocity lớn
+```
+
+Đây là điểm giúp model chuyên biệt cho fall detection.
+
+---
+
+# 11. Mô hình DynaFall-GCN
+
+## Kiến trúc đề xuất
+
+```text
+Input skeleton sequence
+    ↓
+Confidence-aware keypoint dropout
+    ↓
+Joint GCN branch
+    ↓
+Bone GCN branch
+    ↓
+Dynamics temporal branch
+    ↓
+Feature fusion
+    ↓
+MLP classifier
+    ↓
+Fall / Non-fall
+```
+
+## Chi tiết 3 branch
+
+### Branch 1: Joint GCN
+
+Học quan hệ giữa các khớp.
+
+```text
+Input: normalized joint coordinates
+Encoder: lightweight ST-GCN / graph-temporal block
+```
+
+### Branch 2: Bone GCN
+
+Học hướng và độ thay đổi của xương.
+
+```text
+Input: bone vectors
+Encoder: graph-temporal block
+```
+
+### Branch 3: Dynamics branch
+
+Học các đặc trưng rơi.
+
+```text
+Input: velocity, acceleration, torso angle, hip drop, aspect ratio
+Encoder: TCN + attention pooling
+```
+
+### Fusion
+
+```text
+f = concat(f_joint, f_bone, f_dyn)
+logits = MLP(f)
+```
+
+---
+
+# 12. Confidence-aware keypoint dropout
+
+Đây là điểm mới dễ viết.
+
+## Ý tưởng
+
+Trong thực tế, pose estimator có thể mất keypoint do:
+
+```text
+che khuất
+camera góc xấu
+người nằm sát sàn
+motion blur
+```
+
+Do đó, khi train, mình chủ động mask keypoints.
+
+## Cách làm
+
+Nếu confidence thấp:
+
+```text
+conf < 0.3
+```
+
+thì xác suất bị mask cao hơn.
+
+Ví dụ:
+
+```text
+high-confidence keypoint: mask probability = 0.1
+low-confidence keypoint: mask probability = 0.5
+```
+
+Khi mask:
+
+```text
+x = 0
+y = 0
+confidence = 0
+velocity = 0
+acceleration = 0
+```
+
+---
+
+# 13. Thí nghiệm cần chạy
+
+## Experiment 1: Main comparison
+
+Chạy tất cả method trên từng dataset.
+
+```text
+URFD train/test
+MCFD train/test
+```
+
+Bảng:
+
+| Method           | URFD Acc | URFD Recall | URFD F1 | MCFD Acc | MCFD Recall | MCFD F1 |
+| ---------------- | -------: | ----------: | ------: | -------: | ----------: | ------: |
+| LSTM             |          |             |         |          |             |         |
+| ST-GCN           |          |             |         |          |             |         |
+| 2s-AGCN          |          |             |         |          |             |         |
+| CTR-GCN          |          |             |         |          |             |         |
+| PoseC3D          |          |             |         |          |             |         |
+| TCNTE-style      |          |             |         |          |             |         |
+| **DynaFall-GCN** |          |             |         |          |             |         |
+
+---
+
+## Experiment 2: Ablation study
+
+Chỉ chạy trên model của mình.
+
+| Variant | Joint | Bone | Dynamics | Dropout                  | F1 | Recall |
+| ------- | ----- | ---- | -------- | ------------------------ | -: | -----: |
+| A1      | ✓     |      |          |                          |    |        |
+| A2      | ✓     | ✓    |          |                          |    |        |
+| A3      | ✓     | ✓    | ✓        |                          |    |        |
+| A4      | ✓     | ✓    | ✓        | Random dropout           |    |        |
+| A5      | ✓     | ✓    | ✓        | Confidence-aware dropout |    |        |
+
+Mục tiêu:
+
+```text
+Bone giúp tăng F1.
+Dynamics giúp tăng Recall.
+Confidence-aware dropout giúp robustness tốt hơn.
+```
+
+---
+
+## Experiment 3: Robustness under missing keypoints
+
+Tạo test set bị mất keypoint:
+
+```text
+Clean
+Missing-10
+Missing-20
+Missing-30
+Upper-body missing
+Lower-body missing
+Low-confidence masking
+```
+
+Bảng:
+
+| Method           | Clean F1 | Missing-10 | Missing-20 | Missing-30 | Lower-body missing |
+| ---------------- | -------: | ---------: | ---------: | ---------: | -----------------: |
+| ST-GCN           |          |            |            |            |                    |
+| 2s-AGCN          |          |            |            |            |                    |
+| CTR-GCN          |          |            |            |            |                    |
+| PoseC3D          |          |            |            |            |                    |
+| TCNTE-style      |          |            |            |            |                    |
+| **DynaFall-GCN** |          |            |            |            |                    |
+
+Đây là bảng nên nhấn mạnh trong paper.
+
+---
+
+## Experiment 4: Cross-dataset generalization
+
+```text
+Train URFD → Test MCFD
+Train MCFD → Test URFD
+```
+
+Bảng:
+
+| Train | Test | ST-GCN | 2s-AGCN | CTR-GCN | PoseC3D | TCNTE-style | Ours |
+| ----- | ---- | -----: | ------: | ------: | ------: | ----------: | ---: |
+| URFD  | MCFD |        |         |         |         |             |      |
+| MCFD  | URFD |        |         |         |         |             |      |
+
+Nếu kết quả thấp cũng không sao. Cross-dataset thường khó. Chỉ cần proposed giảm ít hơn baseline.
+
+---
+
+## Experiment 5: Runtime / Efficiency
+
+Báo cáo:
+
+```text
+Parameters
+FLOPs nếu tính được
+Inference time per clip
+FPS
+```
+
+Bảng:
+
+| Method       | Params | FPS | F1 |
+| ------------ | -----: | --: | -: |
+| ST-GCN       |        |     |    |
+| CTR-GCN      |        |     |    |
+| PoseC3D      |        |     |    |
+| TCNTE-style  |        |     |    |
+| DynaFall-GCN |        |     |    |
+
+---
+
+# 14. Metrics bắt buộc
+
+Dùng:
+
+```text
+Accuracy
+Precision
+Recall / Sensitivity
+Specificity
+F1-score
+Macro-F1
+FPS
+```
+
+Trong fall detection, cần nhấn mạnh:
+
+```text
+Recall quan trọng hơn Accuracy vì bỏ sót fall nguy hiểm hơn báo nhầm.
+```
+
+---
+
+# 15. Training setup
+
+## Split
+
+Nếu dataset nhỏ:
+
+```text
+5-fold cross-validation
+```
+
+Nếu muốn nhanh:
+
+```text
+70% train
+15% validation
+15% test
+```
+
+Nhưng phải là:
+
+```text
+video-level split
+```
+
+không phải frame-level split.
+
+## Seeds
+
+Tối thiểu:
+
+```text
+7, 13, 21
+```
+
+Tốt hơn:
+
+```text
+7, 13, 21, 42, 100
+```
+
+## Hyperparameters gợi ý
+
+| Setting        |                              Value |
+| -------------- | ---------------------------------: |
+| Clip length    |                                 32 |
+| Stride         |                                 16 |
+| Batch size     |                         16 hoặc 32 |
+| Epochs         |                             80–100 |
+| Optimizer      |                              AdamW |
+| Learning rate  |                     1e-3 hoặc 5e-4 |
+| Weight decay   |                               1e-4 |
+| Early stopping |                        patience 10 |
+| Loss           | CrossEntropy / Weighted Focal Loss |
+
+---
+
+# 16. Cấu trúc paper 8 trang
+
+## Title
+
+**DynaFall: Robust Skeleton-Based Fall Detection via Keypoint Dynamics and Graph-Based Benchmarking**
+
+## Abstract
+
+Nêu 5 ý:
+
+```text
+fall detection quan trọng
+RGB-based methods có privacy/computation issue
+skeleton-based methods nhẹ hơn nhưng dễ bị pose noise
+paper benchmark các skeleton methods mạnh
+đề xuất DynaFall-GCN với dynamics + confidence-aware dropout
+```
+
+## 1. Introduction
+
+Flow:
+
+```text
+Fall detection quan trọng cho elderly care
+Wearable sensors bất tiện, RGB video có privacy concern
+Skeleton/keypoints là hướng nhẹ và privacy-preserving
+Gap: nhiều bài chỉ so baseline đơn giản, ít benchmark method GCN/PoseC3D mạnh trên fall datasets
+Gap: ít kiểm thử robustness khi mất keypoint
+Contribution
+```
+
+## 2. Related Work
+
+Chia 4 phần:
+
+```text
+2.1 Vision-based fall detection
+2.2 Skeleton-based action recognition
+2.3 Graph-based skeleton models
+2.4 Robust fall detection under occlusion
+```
+
+## 3. Proposed Method
+
+```text
+3.1 Overview
+3.2 Pose extraction
+3.3 Skeleton normalization
+3.4 Joint, bone, and dynamics representation
+3.5 Confidence-aware keypoint dropout
+3.6 DynaFall-GCN architecture
+```
+
+## 4. Experiments
+
+```text
+4.1 Datasets
+4.2 Comparative methods
+4.3 Evaluation protocol
+4.4 Implementation details
+4.5 Metrics
+```
+
+## 5. Results and Discussion
+
+```text
+5.1 Main comparison
+5.2 Ablation study
+5.3 Robustness analysis
+5.4 Cross-dataset generalization
+5.5 Runtime analysis
+5.6 Error analysis
+```
+
+## 6. Conclusion
+
+Nêu:
+
+```text
+DynaFall-GCN hiệu quả và robust
+skeleton giúp privacy-preserving
+hạn chế: dataset vẫn là simulated falls
+future work: real-world elderly care, edge deployment, multimodal fusion
+```
+
+---
+
+# 17. Figures nên có
+
+## Figure 1: Overall framework
+
+```text
+RGB Video
+→ Pose Extraction
+→ Skeleton Normalization
+→ Joint/Bone/Dynamics Features
+→ Comparative Skeleton Methods
+→ DynaFall-GCN
+→ Fall / Non-fall
+```
+
+## Figure 2: DynaFall-GCN architecture
+
+```text
+Joint Stream
+Bone Stream
+Dynamics Stream
+→ Fusion
+→ Classifier
+```
+
+## Figure 3: Robustness scenarios
+
+```text
+Clean skeleton
+Missing-10
+Missing-20
+Missing-30
+Lower-body occlusion
+```
+
+---
+
+# 18. Bảng cần có trong paper
+
+Tối thiểu nên có 5 bảng:
+
+```text
+Table 1. Dataset statistics
+Table 2. Main comparison with related methods
+Table 3. Ablation study
+Table 4. Robustness under missing keypoints
+Table 5. Cross-dataset generalization
+```
+
+Nếu thiếu trang thì bỏ Table 5 hoặc đưa vào phụ lục, nhưng **không nên bỏ Table 3 và Table 4**.
+
+---
+
+# 19. Lộ trình chạy thực nghiệm
+
+## Phase 1: Dataset + pose extraction
+
+Làm:
+
+```text
+Download URFD + MCFD
+Extract frames
+Run YOLO pose
+Save keypoints
+Visual check skeleton
+```
+
+Output:
+
+```text
+URFD_keypoints.pkl
+MCFD_keypoints.pkl
+```
+
+---
+
+## Phase 2: Build clips
+
+Làm:
+
+```text
+video-level split
+32-frame clips
+stride 16
+normalize skeleton
+generate joint/bone/dynamics features
+```
+
+Output:
+
+```text
+URFD_train.pkl / val.pkl / test.pkl
+MCFD_train.pkl / val.pkl / test.pkl
+```
+
+---
+
+## Phase 3: Chạy baseline dễ trước
+
+Chạy:
+
+```text
+LSTM
+ST-GCN
+```
+
+Mục tiêu:
+
+```text
+kiểm tra data pipeline đúng chưa
+```
+
+---
+
+## Phase 4: Chạy method từ paper khác
+
+Chạy:
+
+```text
+2s-AGCN
+CTR-GCN
+PoseC3D
+TCNTE-style
+```
+
+---
+
+## Phase 5: Chạy proposed
+
+Chạy:
+
+```text
+DynaFall-GCN joint only
+DynaFall-GCN joint + bone
+DynaFall-GCN joint + bone + dynamics
+DynaFall-GCN full + confidence-aware dropout
+```
+
+---
+
+## Phase 6: Robustness test
+
+Test tất cả model chính trên:
+
+```text
+Clean
+Missing-10
+Missing-20
+Missing-30
+Lower-body missing
+Low-confidence masking
+```
+
+---
+
+## Phase 7: Viết paper
+
+Thứ tự viết nên là:
+
+```text
+Method
+Experiments
+Results
+Introduction
+Related Work
+Abstract
+Conclusion
+```
+
+---
+
+# 20. Bản tối giản nếu thiếu thời gian
+
+Nếu deadline gấp, chỉ làm:
+
+```text
+Datasets:
+URFD + MCFD
+
+Methods:
+LSTM
+ST-GCN
+2s-AGCN
+CTR-GCN
+TCNTE-style
+DynaFall-GCN
+
+Bỏ nếu không kịp:
+PoseC3D
+Cross-dataset
+5 seeds
+```
+
+Nhưng giữ bắt buộc:
+
+```text
+Main comparison
+Ablation
+Robustness missing keypoints
+```
+
+---
+
+# 21. Kế hoạch cuối cùng nên theo
+
+Chốt hướng như sau:
+
+```text
+Paper title:
+DynaFall: Robust Skeleton-Based Fall Detection via Keypoint Dynamics and Graph-Based Benchmarking
+
+Datasets:
+URFD + MCFD
+UP-Fall optional
+
+Pose extractor:
+YOLOv8n-pose hoặc YOLO11n-pose
+
+Compared methods:
+LSTM
+ST-GCN
+2s-AGCN
+CTR-GCN
+PoseC3D nếu kịp
+TCNTE-style
+
+Proposed:
+DynaFall-GCN
+= joint stream
++ bone stream
++ fall-specific dynamics stream
++ confidence-aware keypoint dropout
+
+Experiments:
+Main results
+Ablation
+Robustness
+Cross-dataset nếu kịp
+Runtime/FPS
+
+Main claim:
+DynaFall-GCN is a lightweight and privacy-preserving skeleton-based fall detection model that achieves competitive performance and better robustness under missing or unreliable keypoints.
+```
+
+Đây là kế hoạch hợp lý nhất cho hội nghị Q4: **đủ mới, đủ method từ paper khác, đủ thí nghiệm, nhưng vẫn chạy được trong thời gian ngắn**.
+
+[1]: https://fenix.ur.edu.pl/~mkepski/ds/uf.html?utm_source=chatgpt.com "UR Fall Detection Dataset"
+[2]: https://www.iro.umontreal.ca/~labimage/Dataset/?utm_source=chatgpt.com "Multiple cameras fall dataset"
+[3]: https://sites.google.com/up.edu.mx/challenge-up-2019/data?utm_source=chatgpt.com "Challenge UP - Data"
+[4]: https://www.mdpi.com/1424-8220/19/9/1988?utm_source=chatgpt.com "UP-Fall Detection Dataset: A Multimodal Approach"
+[5]: https://ojs.aaai.org/index.php/aaai/article/view/12328?utm_source=chatgpt.com "Spatial Temporal Graph Convolutional Networks ..."
+[6]: https://arxiv.org/abs/1805.07694?utm_source=chatgpt.com "Two-Stream Adaptive Graph Convolutional Networks for Skeleton-Based Action Recognition"
+[7]: https://arxiv.org/abs/2107.12213?utm_source=chatgpt.com "Channel-wise Topology Refinement Graph Convolution for Skeleton-Based Action Recognition"
+[8]: https://openaccess.thecvf.com/content/CVPR2022/html/Duan_Revisiting_Skeleton-Based_Action_Recognition_CVPR_2022_paper.html?utm_source=chatgpt.com "Related Material - CVPR 2022 Open Access Repository"
+[9]: https://www.sciencedirect.com/science/article/abs/pii/S1574119225000057?utm_source=chatgpt.com "A Real-time skeleton-based fall detection algorithm ..."

PLAN.md

@@ -0,0 +1,1175 @@
+Dưới đây là **kế hoạch hoàn chỉnh phiên bản mạnh hơn** cho paper phát hiện té ngã, có dataset link, mô tả dataset, method lấy từ các paper khác, proposed method, thí nghiệm, bảng kết quả và lộ trình chạy.
+
+# 1. Hướng paper chốt
+
+## Tên paper đề xuất
+
+**DynaFall: Robust Skeleton-Based Fall Detection via Keypoint Dynamics and Graph-Based Benchmarking**
+
+Tên này tốt hơn tên cũ vì nó thể hiện 2 điểm:
+
+1. Có **model riêng**: DynaFall.
+2. Có **benchmark với các method mạnh từ paper khác**: ST-GCN, 2s-AGCN, CTR-GCN, PoseC3D, TCNTE-style.
+
+Ý tưởng chính:
+
+```text
+RGB Video
+→ YOLO Pose
+→ 17-keypoint skeleton sequence
+→ Joint / Bone / Dynamics features
+→ Compare with strong skeleton methods
+→ Proposed DynaFall-GCN
+→ Fall / Non-fall
+```
+
+Paper sẽ không còn đơn giản là “dùng CNN/LSTM/Transformer để phân loại té ngã”, mà thành:
+
+> Một framework phát hiện té ngã dựa trên skeleton, chạy lại các method skeleton action recognition mạnh trên cùng dataset, sau đó đề xuất model chuyên biệt cho fall detection bằng keypoint dynamics và confidence-aware dropout.
+
+---
+
+# 2. Dataset dùng trong paper
+
+Tôi khuyên dùng **2 dataset chính + 1 dataset backup/extension**.
+
+## Dataset chính 1: UR Fall Detection Dataset — URFD
+
+**Link dataset:** UR Fall Detection Dataset. ([Fenix][1])
+
+**Mô tả:** URFD gồm **70 sequences**, trong đó có **30 fall sequences** và **40 activities of daily living — ADL sequences**. Fall events được ghi bằng **2 Microsoft Kinect cameras** kèm dữ liệu accelerometer; ADL events được ghi bằng camera 0 và accelerometer. ([Fenix][1])
+
+**Vì sao nên dùng:**
+
+* Nhỏ, dễ chạy nhanh.
+* Rất phổ biến trong fall detection.
+* Phù hợp để debug pipeline trước.
+* Có RGB video để trích xuất YOLO pose.
+
+**Vai trò trong paper:**
+
+```text
+Dataset kiểm tra nhanh + benchmark chính.
+```
+
+---
+
+## Dataset chính 2: Multiple Cameras Fall Dataset — MCFD
+
+**Link dataset:** Multiple Cameras Fall Dataset. ([Iro Montéal][2])
+
+**Mô tả:** MCFD có **24 scenarios** được ghi bằng **8 IP video cameras**. **22 scenarios đầu** chứa fall và các sự kiện gây nhầm lẫn; **2 scenarios cuối** chỉ chứa các sự kiện gây nhầm lẫn. ([Iro Montéal][2])
+
+**Vì sao nên dùng:**
+
+* Có nhiều góc camera.
+* Phù hợp để kiểm tra robustness.
+* Paper nhìn mạnh hơn URFD-only.
+* Có các tình huống dễ gây nhầm: ngồi, cúi, nằm, chuyển động gần giống fall.
+
+**Vai trò trong paper:**
+
+```text
+Dataset kiểm tra multi-view + robustness.
+```
+
+---
+
+## Dataset backup/extension: UP-Fall Detection Dataset
+
+**Link dataset:** UP-Fall / Challenge UP data page. ([Google Sites][3])
+
+**Mô tả:** Trang Challenge UP mô tả UP-Fall là dataset lớn cho fall detection, gồm **11 activities**, **3 trials per activity**, thực hiện bởi **12 subjects**, bao gồm 6 hoạt động hằng ngày và 5 kiểu té ngã, thu bằng wearable sensors, ambient sensors và vision devices. ([Google Sites][3]) Paper gốc về UP-Fall cũng nhấn mạnh đây là dataset multimodal cho fall detection và human activity recognition. ([MDPI][4])
+
+**Khi nào dùng:**
+
+* Nếu MCFD tải khó hoặc xử lý annotation khó.
+* Nếu muốn paper mạnh hơn và có thêm dataset thứ 3.
+* Nếu muốn hướng “multimodal dataset nhưng chỉ dùng RGB-derived pose”.
+
+**Khuyến nghị thực tế:**
+Paper 8 trang cho hội nghị Q4 thì dùng **URFD + MCFD** là đủ. UP-Fall để backup.
+
+---
+
+# 3. Dataset protocol chốt
+
+Trong paper ghi rõ:
+
+```text
+Although some datasets provide depth, accelerometer, or multimodal signals, this study only uses RGB videos to extract human skeleton keypoints. This ensures a consistent vision-based evaluation protocol across all datasets.
+```
+
+Tức là:
+
+| Dataset          | Dữ liệu gốc có gì         | Mình dùng gì               |
+| ---------------- | ------------------------- | -------------------------- |
+| URFD             | RGB, depth, accelerometer | Chỉ RGB → YOLO Pose        |
+| MCFD             | Multi-camera RGB video    | RGB → YOLO Pose            |
+| UP-Fall nếu dùng | wearable, ambient, vision | Chỉ RGB/vision → YOLO Pose |
+
+Điểm này rất quan trọng để reviewer thấy so sánh công bằng.
+
+---
+
+# 4. Proposed method
+
+## Tên model
+
+**DynaFall-GCN**
+
+## Ý tưởng
+
+Model của mình không chỉ dùng skeleton thô, mà có 3 stream:
+
+```text
+Stream 1: Joint stream
+- x, y, confidence của 17 keypoints
+
+Stream 2: Bone stream
+- vector xương giữa các keypoints
+
+Stream 3: Dynamics stream
+- velocity
+- acceleration
+- torso angle
+- hip drop
+- body aspect ratio
+- center-of-body motion
+```
+
+Sau đó dùng:
+
+```text
+Graph-temporal encoder
++ confidence-aware keypoint dropout
++ classification head
+```
+
+---
+
+# 5. Pipeline tổng thể
+
+```text
+Input RGB Video
+    ↓
+Frame Sampling
+T = 32 frames
+    ↓
+YOLOv8/YOLO11 Pose Extraction
+17 COCO keypoints per frame
+    ↓
+Pose Normalization
+bbox-centered normalization
+    ↓
+Feature Construction
+joint + bone + dynamics
+    ↓
+Comparative Methods
+ST-GCN / 2s-AGCN / CTR-GCN / PoseC3D / TCNTE-style
+    ↓
+Proposed DynaFall-GCN
+joint stream + bone stream + dynamics stream
+    ↓
+Robustness Evaluation
+missing keypoints / lower-body occlusion / low-confidence mask
+    ↓
+Fall / Non-fall Prediction
+```
+
+---
+
+# 6. Các method từ paper khác để chạy lại
+
+Đây là phần giúp paper “hay hơn” và không bị đơn giản.
+
+## 6.1 LSTM baseline
+
+Dùng để kiểm tra sequence baseline.
+
+Input:
+
+```text
+32 frames × 17 keypoints × 3
+→ flatten
+→ 32 × 51
+```
+
+Vai trò:
+
+```text
+Basic temporal baseline.
+```
+
+---
+
+## 6.2 ST-GCN
+
+**Paper:** Spatial Temporal Graph Convolutional Networks for Skeleton-Based Action Recognition, AAAI 2018.
+
+ST-GCN mô hình hóa skeleton như một graph không-thời gian, trong đó keypoints là nodes, bones là edges, và mô hình học cả quan hệ không gian giữa khớp và quan hệ thời gian giữa frames. ([AAAI Publications][5])
+
+Áp dụng vào paper:
+
+```text
+Input: N × C × T × V × M
+C = 3
+T = 32
+V = 17
+M = 1
+Output = Fall / Non-fall
+```
+
+Vai trò:
+
+```text
+Classic graph-based skeleton baseline.
+```
+
+---
+
+## 6.3 2s-AGCN
+
+**Paper:** Two-Stream Adaptive Graph Convolutional Networks for Skeleton-Based Action Recognition, CVPR 2019.
+
+2s-AGCN mạnh hơn ST-GCN vì graph topology có thể được học thích nghi, không cố định hoàn toàn. Method này cũng dùng hai stream: **joint stream** và **bone stream**, trong đó bone stream chứa thông tin độ dài và hướng xương. ([arXiv][6])
+
+Áp dụng:
+
+```text
+Joint stream: keypoint coordinates
+Bone stream: bone vectors
+Final prediction: score fusion
+```
+
+Vai trò:
+
+```text
+Adaptive graph + bone information baseline.
+```
+
+---
+
+## 6.4 CTR-GCN
+
+**Paper:** Channel-wise Topology Refinement Graph Convolution for Skeleton-Based Action Recognition, ICCV 2021.
+
+CTR-GCN học graph topology theo từng channel, giúp mô hình học các quan hệ khớp khác nhau cho từng loại feature. Đây là một baseline GCN mạnh cho skeleton action recognition. ([arXiv][7])
+
+Áp dụng:
+
+```text
+Input: joint + bone skeleton sequence
+Graph: COCO 17-keypoint topology
+Output: Fall / Non-fall
+```
+
+Vai trò:
+
+```text
+Strong graph-based baseline.
+```
+
+---
+
+## 6.5 PoseC3D / PoseConv3D
+
+**Paper:** Revisiting Skeleton-based Action Recognition, CVPR 2022.
+
+PoseC3D/PoseConv3D không dùng graph sequence trực tiếp, mà chuyển pose thành **3D heatmap volume**. Paper gốc cho rằng cách này hiệu quả hơn trong học spatiotemporal features, robust hơn với pose estimation noise và generalize tốt hơn trong cross-dataset settings. ([CVF Open Access][8])
+
+Áp dụng:
+
+```text
+YOLO keypoints
+→ 3D pose heatmap volume
+→ 3D CNN
+→ Fall / Non-fall
+```
+
+Vai trò:
+
+```text
+Heatmap-based skeleton baseline.
+```
+
+---
+
+## 6.6 TCNTE-style
+
+**Paper:** Real-time skeleton-based fall detection algorithm based on Temporal Convolutional Networks and Transformer Encoder, 2025.
+
+TCNTE kết hợp **Temporal Convolutional Network** với **Transformer Encoder** cho fall detection thời gian thực, dùng YOLOv8 pose extraction, BoT-SORT tracking, và weighted focal loss để xử lý mất cân bằng lớp. ([ScienceDirect][9])
+
+Áp dụng:
+
+```text
+TCN + Transformer Encoder
+Input: skeleton sequence
+Loss: weighted focal loss
+Output: Fall / Non-fall
+```
+
+Vai trò:
+
+```text
+Recent fall-specific temporal baseline.
+```
+
+---
+
+# 7. Bảng method chính trong paper
+
+Nên đưa bảng này vào phần **Experimental Setup**.
+
+| Method           | Loại              | Nguồn gốc           | Input                   | Vai trò                     |
+| ---------------- | ----------------- | ------------------- | ----------------------- | --------------------------- |
+| LSTM             | Sequence model    | baseline phổ biến   | joint sequence          | baseline cơ bản             |
+| ST-GCN           | Graph-based       | AAAI 2018           | joint graph             | skeleton baseline kinh điển |
+| 2s-AGCN          | Adaptive GCN      | CVPR 2019           | joint + bone            | baseline mạnh hơn ST-GCN    |
+| CTR-GCN          | Refined GCN       | ICCV 2021           | joint + bone            | strong GCN baseline         |
+| PoseC3D          | Heatmap 3D CNN    | CVPR 2022           | pose heatmap            | khác hướng GCN              |
+| TCNTE-style      | TCN + Transformer | fall detection 2025 | skeleton sequence       | fall-specific baseline      |
+| **DynaFall-GCN** | Proposed          | ours                | joint + bone + dynamics | model chính                 |
+
+---
+
+# 8. Điểm mới của DynaFall-GCN so với các method kia
+
+| Thành phần                        | ST-GCN         | 2s-AGCN        | CTR-GCN        | PoseC3D         | TCNTE                     | DynaFall-GCN |
+| --------------------------------- | -------------- | -------------- | -------------- | --------------- | ------------------------- | ------------ |
+| Joint stream                      | Có             | Có             | Có             | Heatmap         | Có                        | Có           |
+| Bone stream                       | Không          | Có             | Có             | Không trực tiếp | Không rõ                  | Có           |
+| Velocity/Acceleration             | Không          | Không          | Không          | Không trực tiếp | Có thể có motion temporal | Có rõ        |
+| Torso angle                       | Không          | Không          | Không          | Không           | Không                     | Có           |
+| Hip drop                          | Không          | Không          | Không          | Không           | Không                     | Có           |
+| Body aspect ratio                 | Không          | Không          | Không          | Không           | Không                     | Có           |
+| Confidence-aware dropout          | Không          | Không          | Không          | Không           | Không rõ                  | Có           |
+| Robustness test missing keypoints | Không bắt buộc | Không bắt buộc | Không bắt buộc | Có thể robust   | Không bắt buộc            | Có           |
+
+Thông điệp chính:
+
+> Các method skeleton action recognition mạnh học action nói chung, còn DynaFall-GCN đưa thêm fall-specific dynamics và robustness training để phù hợp hơn với bài toán té ngã.
+
+---
+
+# 9. Chi tiết xử lý dữ liệu
+
+## Step 1: Download dataset
+
+Tải:
+
+1. **URFD** từ trang chính thức. ([Fenix][1])
+2. **MCFD** từ trang dataset của Université de Montréal. ([Iro Montéal][2])
+3. **UP-Fall** chỉ dùng nếu cần backup hoặc mở rộng. ([Google Sites][3])
+
+---
+
+## Step 2: Chuẩn hóa thư mục
+
+```text
+fall_project/
+├── data/
+│   ├── raw/
+│   │   ├── URFD/
+│   │   ├── MCFD/
+│   │   └── UPFall_optional/
+│   ├── poses/
+│   ├── clips/
+│   └── splits/
+├── src/
+├── configs/
+├── results/
+└── paper/
+```
+
+---
+
+## Step 3: Trích xuất pose
+
+Dùng:
+
+```text
+YOLOv8n-pose hoặc YOLO11n-pose
+```
+
+Output mỗi frame:
+
+```text
+17 keypoints × 3
+x, y, confidence
+```
+
+Nếu có nhiều người trong frame:
+
+```text
+Chọn người có bbox lớn nhất.
+```
+
+Lý do:
+
+```text
+Fall datasets thường chỉ có một subject chính.
+```
+
+---
+
+## Step 4: Normalize keypoints
+
+Với mỗi frame:
+
+```text
+x_norm = (x - bbox_center_x) / bbox_width
+y_norm = (y - bbox_center_y) / bbox_height
+confidence giữ nguyên
+```
+
+Output:
+
+```text
+T × 17 × 3
+```
+
+Trong đó:
+
+```text
+T = 32 frames
+```
+
+---
+
+## Step 5: Tạo clip
+
+Dùng:
+
+```text
+clip length = 32 frames
+stride = 16
+```
+
+Nếu video ngắn:
+
+```text
+uniform sampling hoặc padding frame cuối
+```
+
+Quan trọng:
+
+```text
+Chia train/val/test theo video trước, rồi mới tạo clip.
+Không random clip trực tiếp vì sẽ leak dữ liệu.
+```
+
+---
+
+# 10. Feature engineering cho proposed model
+
+## 10.1 Joint features
+
+```text
+x, y, confidence
+```
+
+Shape:
+
+```text
+32 × 17 × 3
+```
+
+---
+
+## 10.2 Bone features
+
+Dựa trên COCO skeleton edges:
+
+```text
+bone_vector = child_joint - parent_joint
+```
+
+Các cạnh chính:
+
+```text
+nose-eye-ear
+shoulder-elbow-wrist
+shoulder-hip
+hip-knee-ankle
+left-right shoulder
+left-right hip
+```
+
+Shape:
+
+```text
+32 × 17 × 3
+```
+
+---
+
+## 10.3 Dynamics features
+
+Tạo thêm:
+
+```text
+velocity = p_t - p_{t-1}
+acceleration = v_t - v_{t-1}
+torso angle
+hip drop
+body aspect ratio
+center-of-body velocity
+```
+
+Fall thường có:
+
+```text
+torso angle thay đổi nhanh
+hip y-coordinate tụt nhanh
+bbox height/width thay đổi
+center-of-body velocity lớn
+```
+
+Đây là điểm giúp model chuyên biệt cho fall detection.
+
+---
+
+# 11. Mô hình DynaFall-GCN
+
+## Kiến trúc đề xuất
+
+```text
+Input skeleton sequence
+    ↓
+Confidence-aware keypoint dropout
+    ↓
+Joint GCN branch
+    ↓
+Bone GCN branch
+    ↓
+Dynamics temporal branch
+    ↓
+Feature fusion
+    ↓
+MLP classifier
+    ↓
+Fall / Non-fall
+```
+
+## Chi tiết 3 branch
+
+### Branch 1: Joint GCN
+
+Học quan hệ giữa các khớp.
+
+```text
+Input: normalized joint coordinates
+Encoder: lightweight ST-GCN / graph-temporal block
+```
+
+### Branch 2: Bone GCN
+
+Học hướng và độ thay đổi của xương.
+
+```text
+Input: bone vectors
+Encoder: graph-temporal block
+```
+
+### Branch 3: Dynamics branch
+
+Học các đặc trưng rơi.
+
+```text
+Input: velocity, acceleration, torso angle, hip drop, aspect ratio
+Encoder: TCN + attention pooling
+```
+
+### Fusion
+
+```text
+f = concat(f_joint, f_bone, f_dyn)
+logits = MLP(f)
+```
+
+---
+
+# 12. Confidence-aware keypoint dropout
+
+Đây là điểm mới dễ viết.
+
+## Ý tưởng
+
+Trong thực tế, pose estimator có thể mất keypoint do:
+
+```text
+che khuất
+camera góc xấu
+người nằm sát sàn
+motion blur
+```
+
+Do đó, khi train, mình chủ động mask keypoints.
+
+## Cách làm
+
+Nếu confidence thấp:
+
+```text
+conf < 0.3
+```
+
+thì xác suất bị mask cao hơn.
+
+Ví dụ:
+
+```text
+high-confidence keypoint: mask probability = 0.1
+low-confidence keypoint: mask probability = 0.5
+```
+
+Khi mask:
+
+```text
+x = 0
+y = 0
+confidence = 0
+velocity = 0
+acceleration = 0
+```
+
+---
+
+# 13. Thí nghiệm cần chạy
+
+## Experiment 1: Main comparison
+
+Chạy tất cả method trên từng dataset.
+
+```text
+URFD train/test
+MCFD train/test
+```
+
+Bảng:
+
+| Method           | URFD Acc | URFD Recall | URFD F1 | MCFD Acc | MCFD Recall | MCFD F1 |
+| ---------------- | -------: | ----------: | ------: | -------: | ----------: | ------: |
+| LSTM             |          |             |         |          |             |         |
+| ST-GCN           |          |             |         |          |             |         |
+| 2s-AGCN          |          |             |         |          |             |         |
+| CTR-GCN          |          |             |         |          |             |         |
+| PoseC3D          |          |             |         |          |             |         |
+| TCNTE-style      |          |             |         |          |             |         |
+| **DynaFall-GCN** |          |             |         |          |             |         |
+
+---
+
+## Experiment 2: Ablation study
+
+Chỉ chạy trên model của mình.
+
+| Variant | Joint | Bone | Dynamics | Dropout                  | F1 | Recall |
+| ------- | ----- | ---- | -------- | ------------------------ | -: | -----: |
+| A1      | ✓     |      |          |                          |    |        |
+| A2      | ✓     | ✓    |          |                          |    |        |
+| A3      | ✓     | ✓    | ✓        |                          |    |        |
+| A4      | ✓     | ✓    | ✓        | Random dropout           |    |        |
+| A5      | ✓     | ✓    | ✓        | Confidence-aware dropout |    |        |
+
+Mục tiêu:
+
+```text
+Bone giúp tăng F1.
+Dynamics giúp tăng Recall.
+Confidence-aware dropout giúp robustness tốt hơn.
+```
+
+---
+
+## Experiment 3: Robustness under missing keypoints
+
+Tạo test set bị mất keypoint:
+
+```text
+Clean
+Missing-10
+Missing-20
+Missing-30
+Upper-body missing
+Lower-body missing
+Low-confidence masking
+```
+
+Bảng:
+
+| Method           | Clean F1 | Missing-10 | Missing-20 | Missing-30 | Lower-body missing |
+| ---------------- | -------: | ---------: | ---------: | ---------: | -----------------: |
+| ST-GCN           |          |            |            |            |                    |
+| 2s-AGCN          |          |            |            |            |                    |
+| CTR-GCN          |          |            |            |            |                    |
+| PoseC3D          |          |            |            |            |                    |
+| TCNTE-style      |          |            |            |            |                    |
+| **DynaFall-GCN** |          |            |            |            |                    |
+
+Đây là bảng nên nhấn mạnh trong paper.
+
+---
+
+## Experiment 4: Cross-dataset generalization
+
+```text
+Train URFD → Test MCFD
+Train MCFD → Test URFD
+```
+
+Bảng:
+
+| Train | Test | ST-GCN | 2s-AGCN | CTR-GCN | PoseC3D | TCNTE-style | Ours |
+| ----- | ---- | -----: | ------: | ------: | ------: | ----------: | ---: |
+| URFD  | MCFD |        |         |         |         |             |      |
+| MCFD  | URFD |        |         |         |         |             |      |
+
+Nếu kết quả thấp cũng không sao. Cross-dataset thường khó. Chỉ cần proposed giảm ít hơn baseline.
+
+---
+
+## Experiment 5: Runtime / Efficiency
+
+Báo cáo:
+
+```text
+Parameters
+FLOPs nếu tính được
+Inference time per clip
+FPS
+```
+
+Bảng:
+
+| Method       | Params | FPS | F1 |
+| ------------ | -----: | --: | -: |
+| ST-GCN       |        |     |    |
+| CTR-GCN      |        |     |    |
+| PoseC3D      |        |     |    |
+| TCNTE-style  |        |     |    |
+| DynaFall-GCN |        |     |    |
+
+---
+
+# 14. Metrics bắt buộc
+
+Dùng:
+
+```text
+Accuracy
+Precision
+Recall / Sensitivity
+Specificity
+F1-score
+Macro-F1
+FPS
+```
+
+Trong fall detection, cần nhấn mạnh:
+
+```text
+Recall quan trọng hơn Accuracy vì bỏ sót fall nguy hiểm hơn báo nhầm.
+```
+
+---
+
+# 15. Training setup
+
+## Split
+
+Nếu dataset nhỏ:
+
+```text
+5-fold cross-validation
+```
+
+Nếu muốn nhanh:
+
+```text
+70% train
+15% validation
+15% test
+```
+
+Nhưng phải là:
+
+```text
+video-level split
+```
+
+không phải frame-level split.
+
+## Seeds
+
+Tối thiểu:
+
+```text
+7, 13, 21
+```
+
+Tốt hơn:
+
+```text
+7, 13, 21, 42, 100
+```
+
+## Hyperparameters gợi ý
+
+| Setting        |                              Value |
+| -------------- | ---------------------------------: |
+| Clip length    |                                 32 |
+| Stride         |                                 16 |
+| Batch size     |                         16 hoặc 32 |
+| Epochs         |                             80–100 |
+| Optimizer      |                              AdamW |
+| Learning rate  |                     1e-3 hoặc 5e-4 |
+| Weight decay   |                               1e-4 |
+| Early stopping |                        patience 10 |
+| Loss           | CrossEntropy / Weighted Focal Loss |
+
+---
+
+# 16. Cấu trúc paper 8 trang
+
+## Title
+
+**DynaFall: Robust Skeleton-Based Fall Detection via Keypoint Dynamics and Graph-Based Benchmarking**
+
+## Abstract
+
+Nêu 5 ý:
+
+```text
+fall detection quan trọng
+RGB-based methods có privacy/computation issue
+skeleton-based methods nhẹ hơn nhưng dễ bị pose noise
+paper benchmark các skeleton methods mạnh
+đề xuất DynaFall-GCN với dynamics + confidence-aware dropout
+```
+
+## 1. Introduction
+
+Flow:
+
+```text
+Fall detection quan trọng cho elderly care
+Wearable sensors bất tiện, RGB video có privacy concern
+Skeleton/keypoints là hướng nhẹ và privacy-preserving
+Gap: nhiều bài chỉ so baseline đơn giản, ít benchmark method GCN/PoseC3D mạnh trên fall datasets
+Gap: ít kiểm thử robustness khi mất keypoint
+Contribution
+```
+
+## 2. Related Work
+
+Chia 4 phần:
+
+```text
+2.1 Vision-based fall detection
+2.2 Skeleton-based action recognition
+2.3 Graph-based skeleton models
+2.4 Robust fall detection under occlusion
+```
+
+## 3. Proposed Method
+
+```text
+3.1 Overview
+3.2 Pose extraction
+3.3 Skeleton normalization
+3.4 Joint, bone, and dynamics representation
+3.5 Confidence-aware keypoint dropout
+3.6 DynaFall-GCN architecture
+```
+
+## 4. Experiments
+
+```text
+4.1 Datasets
+4.2 Comparative methods
+4.3 Evaluation protocol
+4.4 Implementation details
+4.5 Metrics
+```
+
+## 5. Results and Discussion
+
+```text
+5.1 Main comparison
+5.2 Ablation study
+5.3 Robustness analysis
+5.4 Cross-dataset generalization
+5.5 Runtime analysis
+5.6 Error analysis
+```
+
+## 6. Conclusion
+
+Nêu:
+
+```text
+DynaFall-GCN hiệu quả và robust
+skeleton giúp privacy-preserving
+hạn chế: dataset vẫn là simulated falls
+future work: real-world elderly care, edge deployment, multimodal fusion
+```
+
+---
+
+# 17. Figures nên có
+
+## Figure 1: Overall framework
+
+```text
+RGB Video
+→ Pose Extraction
+→ Skeleton Normalization
+→ Joint/Bone/Dynamics Features
+→ Comparative Skeleton Methods
+→ DynaFall-GCN
+→ Fall / Non-fall
+```
+
+## Figure 2: DynaFall-GCN architecture
+
+```text
+Joint Stream
+Bone Stream
+Dynamics Stream
+→ Fusion
+→ Classifier
+```
+
+## Figure 3: Robustness scenarios
+
+```text
+Clean skeleton
+Missing-10
+Missing-20
+Missing-30
+Lower-body occlusion
+```
+
+---
+
+# 18. Bảng cần có trong paper
+
+Tối thiểu nên có 5 bảng:
+
+```text
+Table 1. Dataset statistics
+Table 2. Main comparison with related methods
+Table 3. Ablation study
+Table 4. Robustness under missing keypoints
+Table 5. Cross-dataset generalization
+```
+
+Nếu thiếu trang thì bỏ Table 5 hoặc đưa vào phụ lục, nhưng **không nên bỏ Table 3 và Table 4**.
+
+---
+
+# 19. Lộ trình chạy thực nghiệm
+
+## Phase 1: Dataset + pose extraction
+
+Làm:
+
+```text
+Download URFD + MCFD
+Extract frames
+Run YOLO pose
+Save keypoints
+Visual check skeleton
+```
+
+Output:
+
+```text
+URFD_keypoints.pkl
+MCFD_keypoints.pkl
+```
+
+---
+
+## Phase 2: Build clips
+
+Làm:
+
+```text
+video-level split
+32-frame clips
+stride 16
+normalize skeleton
+generate joint/bone/dynamics features
+```
+
+Output:
+
+```text
+URFD_train.pkl / val.pkl / test.pkl
+MCFD_train.pkl / val.pkl / test.pkl
+```
+
+---
+
+## Phase 3: Chạy baseline dễ trước
+
+Chạy:
+
+```text
+LSTM
+ST-GCN
+```
+
+Mục tiêu:
+
+```text
+kiểm tra data pipeline đúng chưa
+```
+
+---
+
+## Phase 4: Chạy method từ paper khác
+
+Chạy:
+
+```text
+2s-AGCN
+CTR-GCN
+PoseC3D
+TCNTE-style
+```
+
+---
+
+## Phase 5: Chạy proposed
+
+Chạy:
+
+```text
+DynaFall-GCN joint only
+DynaFall-GCN joint + bone
+DynaFall-GCN joint + bone + dynamics
+DynaFall-GCN full + confidence-aware dropout
+```
+
+---
+
+## Phase 6: Robustness test
+
+Test tất cả model chính trên:
+
+```text
+Clean
+Missing-10
+Missing-20
+Missing-30
+Lower-body missing
+Low-confidence masking
+```
+
+---
+
+## Phase 7: Viết paper
+
+Thứ tự viết nên là:
+
+```text
+Method
+Experiments
+Results
+Introduction
+Related Work
+Abstract
+Conclusion
+```
+
+---
+
+# 20. Bản tối giản nếu thiếu thời gian
+
+Nếu deadline gấp, chỉ làm:
+
+```text
+Datasets:
+URFD + MCFD
+
+Methods:
+LSTM
+ST-GCN
+2s-AGCN
+CTR-GCN
+TCNTE-style
+DynaFall-GCN
+
+Bỏ nếu không kịp:
+PoseC3D
+Cross-dataset
+5 seeds
+```
+
+Nhưng giữ bắt buộc:
+
+```text
+Main comparison
+Ablation
+Robustness missing keypoints
+```
+
+---
+
+# 21. Kế hoạch cuối cùng nên theo
+
+Chốt hướng như sau:
+
+```text
+Paper title:
+DynaFall: Robust Skeleton-Based Fall Detection via Keypoint Dynamics and Graph-Based Benchmarking
+
+Datasets:
+URFD + MCFD
+UP-Fall optional
+
+Pose extractor:
+YOLOv8n-pose hoặc YOLO11n-pose
+
+Compared methods:
+LSTM
+ST-GCN
+2s-AGCN
+CTR-GCN
+PoseC3D nếu kịp
+TCNTE-style
+
+Proposed:
+DynaFall-GCN
+= joint stream
++ bone stream
++ fall-specific dynamics stream
++ confidence-aware keypoint dropout
+
+Experiments:
+Main results
+Ablation
+Robustness
+Cross-dataset nếu kịp
+Runtime/FPS
+
+Main claim:
+DynaFall-GCN is a lightweight and privacy-preserving skeleton-based fall detection model that achieves competitive performance and better robustness under missing or unreliable keypoints.
+```
+
+Đây là kế hoạch hợp lý nhất cho hội nghị Q4: **đủ mới, đủ method từ paper khác, đủ thí nghiệm, nhưng vẫn chạy được trong thời gian ngắn**.
+
+[1]: https://fenix.ur.edu.pl/~mkepski/ds/uf.html?utm_source=chatgpt.com "UR Fall Detection Dataset"
+[2]: https://www.iro.umontreal.ca/~labimage/Dataset/?utm_source=chatgpt.com "Multiple cameras fall dataset"
+[3]: https://sites.google.com/up.edu.mx/challenge-up-2019/data?utm_source=chatgpt.com "Challenge UP - Data"
+[4]: https://www.mdpi.com/1424-8220/19/9/1988?utm_source=chatgpt.com "UP-Fall Detection Dataset: A Multimodal Approach"
+[5]: https://ojs.aaai.org/index.php/aaai/article/view/12328?utm_source=chatgpt.com "Spatial Temporal Graph Convolutional Networks ..."
+[6]: https://arxiv.org/abs/1805.07694?utm_source=chatgpt.com "Two-Stream Adaptive Graph Convolutional Networks for Skeleton-Based Action Recognition"
+[7]: https://arxiv.org/abs/2107.12213?utm_source=chatgpt.com "Channel-wise Topology Refinement Graph Convolution for Skeleton-Based Action Recognition"
+[8]: https://openaccess.thecvf.com/content/CVPR2022/html/Duan_Revisiting_Skeleton-Based_Action_Recognition_CVPR_2022_paper.html?utm_source=chatgpt.com "Related Material - CVPR 2022 Open Access Repository"
+[9]: https://www.sciencedirect.com/science/article/abs/pii/S1574119225000057?utm_source=chatgpt.com "A Real-time skeleton-based fall detection algorithm ..."

README.md

@@ -0,0 +1,58 @@
+# DynaFall experiments
+
+This repository implements the experiment plan in `PLAN.md` for skeleton-based fall detection.
+
+## Structure
+
+```text
+data/raw/              Raw videos: data/raw/URFD and data/raw/MCFD
+data/poses/            Extracted YOLO pose files
+data/processed/        Video-level splits and 32-frame clips
+configs/default.yaml   Main experiment config
+src/dynafall/          Dataset, features, models, training, evaluation
+scripts/               Entry-point scripts
+results/               Metrics, checkpoints, tables
+```
+
+## Quick smoke test
+
+```bash
+python scripts/make_synthetic_dataset.py --dataset Synthetic --videos 24
+python scripts/prepare_clips.py --dataset Synthetic
+python scripts/train.py --dataset Synthetic --method dynafall --epochs 2
+python scripts/evaluate.py --dataset Synthetic --method dynafall
+```
+
+## Real data workflow
+
+Place videos under:
+
+```text
+data/raw/URFD/fall/*.avi
+data/raw/URFD/nonfall/*.avi
+data/raw/MCFD/fall/*.avi
+data/raw/MCFD/nonfall/*.avi
+```
+
+Any common video extension is accepted. Labels are inferred from the parent directory name:
+`fall`, `falls`, `1`, `positive` map to fall; all other directory names map to non-fall.
+
+Then run:
+
+```bash
+python scripts/extract_pose.py --dataset URFD
+python scripts/prepare_clips.py --dataset URFD
+python scripts/run_experiments.py --dataset URFD --methods lstm stgcn agcn ctrgcn posec3d tcnte dynafall
+python scripts/robustness.py --dataset URFD --methods stgcn agcn ctrgcn posec3d tcnte dynafall
+python scripts/aggregate_results.py
+```
+
+Repeat for `MCFD`. Cross-dataset evaluation:
+
+```bash
+python scripts/evaluate.py --dataset MCFD --method dynafall --checkpoint results/URFD/dynafall/best.pt --tag trainURFD_testMCFD
+```
+
+## Notes
+
+The graph baselines are compact reimplementations designed for small fall datasets and a COCO-17 pose layout. They preserve the paper-level comparison categories: LSTM, ST-GCN-style graph temporal model, two-stream adaptive GCN, CTR-GCN-style channel topology refinement, PoseC3D-style heatmap volume, TCN+Transformer, and DynaFall-GCN.

configs/default.yaml

@@ -0,0 +1,28 @@
+seed: 7
+clip_len: 32
+stride: 16
+batch_size: 16
+epochs: 80
+patience: 10
+lr: 0.001
+weight_decay: 0.0001
+num_workers: 2
+device: auto
+pose_model: yolov8n-pose.pt
+conf_threshold: 0.25
+splits:
+  train: 0.7
+  val: 0.15
+  test: 0.15
+model:
+  hidden: 96
+  dropout: 0.25
+  num_classes: 2
+loss:
+  name: focal
+  gamma: 2.0
+dropout:
+  enabled: true
+  high_conf_prob: 0.10
+  low_conf_prob: 0.50
+  low_conf_threshold: 0.30

paper/.ipynb_checkpoints/main-checkpoint.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1e12f6ac51d492ec90cfe8c6b93c2562ad2499f9151345d9c95897eec159a6fc
+size 256352

paper/README.md

@@ -0,0 +1,18 @@
+# Paper draft
+
+This folder contains an anonymous Springer LNCS-style draft.
+
+Files:
+- `main.tex`: anonymous paper draft.
+- `references.bib`: bibliography source kept for reference.
+
+The manuscript intentionally does not include author names, affiliations, email addresses, or acknowledgments to satisfy double-blind review.
+
+Compile with the official Springer LNCS/Proceedings LaTeX package linked from the ICTA submission page. The submitted manuscript uses an inline `thebibliography` block so references remain numbered in citation order:
+
+```bash
+pdflatex main
+pdflatex main
+```
+
+If `llncs.cls` is missing locally, download the official Springer proceedings LaTeX package and place it in this directory or in the TeX search path.

paper/main.aux

@@ -0,0 +1,57 @@
+\relax 
+\citation{alam2022review}
+\providecommand \oddpage@label [2]{}
+\@writefile{toc}{\contentsline {title}{DynaFall: Robust Skeleton-Based Fall Detection via Keypoint Dynamics and Graph-Based Benchmarking}{1}{}\protected@file@percent }
+\@writefile{toc}{\authcount {1}}
+\@writefile{toc}{\contentsline {author}{Anonymous Submission}{1}{}\protected@file@percent }
+\@writefile{toc}{\contentsline {section}{\numberline {1}Introduction}{1}{}\protected@file@percent }
+\citation{urfd}
+\citation{mcfd}
+\citation{alam2022review}
+\citation{urfd}
+\citation{mcfd}
+\@writefile{lof}{\contentsline {figure}{\numberline {1}{\ignorespaces Overall RGB-to-skeleton fall-detection pipeline used for all experiments.}}{2}{}\protected@file@percent }
+\newlabel{fig:pipeline}{{1}{2}{}{}{}}
+\@writefile{toc}{\contentsline {section}{\numberline {2}Related Work}{2}{}\protected@file@percent }
+\citation{yan2018stgcn}
+\citation{shi2019agcn}
+\citation{chen2021ctrgcn}
+\citation{duan2022posec3d}
+\citation{yu2025tcnte}
+\citation{yolov8}
+\@writefile{toc}{\contentsline {section}{\numberline {3}Method}{3}{}\protected@file@percent }
+\@writefile{toc}{\contentsline {subsection}{\numberline {3.1}Pose Extraction and Normalization}{3}{}\protected@file@percent }
+\citation{lin2017focal}
+\@writefile{toc}{\contentsline {subsection}{\numberline {3.2}Feature Construction}{4}{}\protected@file@percent }
+\@writefile{toc}{\contentsline {subsection}{\numberline {3.3}DynaFall-GCN}{4}{}\protected@file@percent }
+\@writefile{toc}{\contentsline {subsection}{\numberline {3.4}Confidence-Aware Keypoint Dropout}{4}{}\protected@file@percent }
+\@writefile{toc}{\contentsline {section}{\numberline {4}Experimental Setup}{4}{}\protected@file@percent }
+\@writefile{toc}{\contentsline {subsection}{\numberline {4.1}Datasets, Splits, and Metrics}{4}{}\protected@file@percent }
+\@writefile{lof}{\contentsline {figure}{\numberline {2}{\ignorespaces DynaFall-GCN architecture with joint, bone, and fall-specific dynamics streams.}}{5}{}\protected@file@percent }
+\newlabel{fig:architecture}{{2}{5}{}{}{}}
+\@writefile{lot}{\contentsline {table}{\numberline {1}{\ignorespaces Dataset statistics after RGB video collection and pose extraction.}}{5}{}\protected@file@percent }
+\newlabel{tab:datasets}{{1}{5}{}{}{}}
+\@writefile{toc}{\contentsline {subsection}{\numberline {4.2}Compared Methods}{5}{}\protected@file@percent }
+\@writefile{lot}{\contentsline {table}{\numberline {2}{\ignorespaces Main comparison over three seeds. Precision, recall, and F1 are computed for the fall class.}}{6}{}\protected@file@percent }
+\newlabel{tab:main}{{2}{6}{}{}{}}
+\@writefile{toc}{\contentsline {section}{\numberline {5}Results}{6}{}\protected@file@percent }
+\@writefile{toc}{\contentsline {subsection}{\numberline {5.1}Main Comparison}{6}{}\protected@file@percent }
+\@writefile{toc}{\contentsline {subsection}{\numberline {5.2}Ablation Study}{6}{}\protected@file@percent }
+\@writefile{lot}{\contentsline {table}{\numberline {3}{\ignorespaces DynaFall-GCN ablation over three seeds. Precision, recall, and F1 are fall-class metrics.}}{7}{}\protected@file@percent }
+\newlabel{tab:ablation}{{3}{7}{}{}{}}
+\@writefile{lof}{\contentsline {figure}{\numberline {3}{\ignorespaces Runtime and efficiency summary. Left: mean fall-class F1 across URFD and MCFD. Right: inference throughput in clips/s.}}{7}{}\protected@file@percent }
+\newlabel{fig:runtime}{{3}{7}{}{}{}}
+\@writefile{toc}{\contentsline {subsection}{\numberline {5.3}Runtime and Efficiency}{7}{}\protected@file@percent }
+\@writefile{toc}{\contentsline {section}{\numberline {6}Discussion}{7}{}\protected@file@percent }
+\bibcite{alam2022review}{1}
+\bibcite{urfd}{2}
+\bibcite{mcfd}{3}
+\bibcite{yan2018stgcn}{4}
+\bibcite{shi2019agcn}{5}
+\bibcite{chen2021ctrgcn}{6}
+\bibcite{duan2022posec3d}{7}
+\bibcite{yu2025tcnte}{8}
+\bibcite{yolov8}{9}
+\bibcite{lin2017focal}{10}
+\@writefile{toc}{\contentsline {section}{\numberline {7}Conclusion}{8}{}\protected@file@percent }
+\gdef \@abspage@last{8}

paper/main.bbl

@@ -0,0 +1,45 @@
+\begin{thebibliography}{1}
+\providecommand{\url}[1]{\texttt{#1}}
+\providecommand{\urlprefix}{URL }
+\providecommand{\doi}[1]{https://doi.org/#1}
+
+\bibitem{mcfd}
+Auvinet, E., Multon, F., Saint-Arnaud, A., Rousseau, J., Meunier, J.: Fall
+  detection using body geometry and human motion analysis. In: International
+  Conference on Aging, Disability and Independence (2011)
+
+\bibitem{chen2021ctrgcn}
+Chen, Y., Zhang, Z., Yuan, C., Li, B., Deng, Y., Hu, W.: Channel-wise topology
+  refinement graph convolution for skeleton-based action recognition. In:
+  IEEE/CVF International Conference on Computer Vision (2021)
+
+\bibitem{duan2022posec3d}
+Duan, H., Zhao, Y., Chen, K., Shao, D., Lin, D., Dai, B.: Revisiting
+  skeleton-based action recognition. In: IEEE/CVF Conference on Computer Vision
+  and Pattern Recognition (2022)
+
+\bibitem{urfd}
+Kepski, M., Kwolek, B.: Human fall detection on embedded platform using depth
+  maps and wireless accelerometer. Computer Methods and Programs in Biomedicine
+   \textbf{117}(3),  489--501 (2014)
+
+\bibitem{lin2017focal}
+Lin, T.Y., Goyal, P., Girshick, R., He, K., Dollar, P.: Focal loss for dense
+  object detection. IEEE Transactions on Pattern Analysis and Machine
+  Intelligence  \textbf{42}(2),  318--327 (2020)
+
+\bibitem{shi2019agcn}
+Shi, L., Zhang, Y., Cheng, J., Lu, H.: Two-stream adaptive graph convolutional
+  networks for skeleton-based action recognition. In: IEEE/CVF Conference on
+  Computer Vision and Pattern Recognition (2019)
+
+\bibitem{yolov8}
+{Ultralytics}: Ultralytics yolov8.
+  \url{https://github.com/ultralytics/ultralytics} (2023)
+
+\bibitem{yan2018stgcn}
+Yan, S., Xiong, Y., Lin, D.: Spatial temporal graph convolutional networks for
+  skeleton-based action recognition. In: AAAI Conference on Artificial
+  Intelligence (2018)
+
+\end{thebibliography}

paper/main.blg

@@ -0,0 +1,46 @@
+This is BibTeX, Version 0.99d (TeX Live 2023/Debian)
+Capacity: max_strings=200000, hash_size=200000, hash_prime=170003
+The top-level auxiliary file: main.aux
+The style file: splncs04.bst
+Database file #1: references.bib
+You've used 8 entries,
+            2850 wiz_defined-function locations,
+            637 strings with 5990 characters,
+and the built_in function-call counts, 4333 in all, are:
+= -- 326
+> -- 176
+< -- 4
++ -- 72
+- -- 64
+* -- 272
+:= -- 581
+add.period$ -- 8
+call.type$ -- 8
+change.case$ -- 66
+chr.to.int$ -- 0
+cite$ -- 8
+duplicate$ -- 385
+empty$ -- 392
+format.name$ -- 71
+if$ -- 919
+int.to.chr$ -- 0
+int.to.str$ -- 8
+missing$ -- 82
+newline$ -- 30
+num.names$ -- 16
+pop$ -- 179
+preamble$ -- 1
+purify$ -- 48
+quote$ -- 0
+skip$ -- 100
+stack$ -- 0
+substring$ -- 168
+swap$ -- 195
+text.length$ -- 4
+text.prefix$ -- 0
+top$ -- 0
+type$ -- 32
+warning$ -- 0
+while$ -- 30
+width$ -- 9
+write$ -- 79

paper/main.log

@@ -0,0 +1,732 @@
+This is pdfTeX, Version 3.141592653-2.6-1.40.25 (TeX Live 2023/Debian) (preloaded format=pdflatex 2026.7.3)  3 JUL 2026 11:07
+entering extended mode
+ restricted \write18 enabled.
+ %&-line parsing enabled.
+**main.tex
+(./main.tex
+LaTeX2e <2023-11-01> patch level 1
+L3 programming layer <2024-01-22>
+(/usr/share/texlive/texmf-dist/tex/latex/llncs/llncs.cls
+Document Class: llncs 2024/01/29 v2.24 
+ LaTeX document class for Lecture Notes in Computer Science
+(/usr/share/texlive/texmf-dist/tex/latex/base/article.cls
+Document Class: article 2023/05/17 v1.4n Standard LaTeX document class
+(/usr/share/texlive/texmf-dist/tex/latex/base/size10.clo
+File: size10.clo 2023/05/17 v1.4n Standard LaTeX file (size option)
+)
+\c@part=\count187
+\c@section=\count188
+\c@subsection=\count189
+\c@subsubsection=\count190
+\c@paragraph=\count191
+\c@subparagraph=\count192
+\c@figure=\count193
+\c@table=\count194
+\abovecaptionskip=\skip48
+\belowcaptionskip=\skip49
+\bibindent=\dimen140
+)
+(/usr/share/texlive/texmf-dist/tex/latex/tools/multicol.sty
+Package: multicol 2023/03/30 v1.9f multicolumn formatting (FMi)
+\c@tracingmulticols=\count195
+\mult@box=\box51
+\multicol@leftmargin=\dimen141
+\c@unbalance=\count196
+\c@collectmore=\count197
+\doublecol@number=\count198
+\multicoltolerance=\count199
+\multicolpretolerance=\count266
+\full@width=\dimen142
+\page@free=\dimen143
+\premulticols=\dimen144
+\postmulticols=\dimen145
+\multicolsep=\skip50
+\multicolbaselineskip=\skip51
+\partial@page=\box52
+\last@line=\box53
+\mc@boxedresult=\box54
+\maxbalancingoverflow=\dimen146
+\mult@rightbox=\box55
+\mult@grightbox=\box56
+\mult@firstbox=\box57
+\mult@gfirstbox=\box58
+\@tempa=\box59
+\@tempa=\box60
+\@tempa=\box61
+\@tempa=\box62
+\@tempa=\box63
+\@tempa=\box64
+\@tempa=\box65
+\@tempa=\box66
+\@tempa=\box67
+\@tempa=\box68
+\@tempa=\box69
+\@tempa=\box70
+\@tempa=\box71
+\@tempa=\box72
+\@tempa=\box73
+\@tempa=\box74
+\@tempa=\box75
+\@tempa=\box76
+\@tempa=\box77
+\@tempa=\box78
+\@tempa=\box79
+\@tempa=\box80
+\@tempa=\box81
+\@tempa=\box82
+\@tempa=\box83
+\@tempa=\box84
+\@tempa=\box85
+\@tempa=\box86
+\@tempa=\box87
+\@tempa=\box88
+\@tempa=\box89
+\@tempa=\box90
+\@tempa=\box91
+\@tempa=\box92
+\@tempa=\box93
+\@tempa=\box94
+\c@minrows=\count267
+\c@columnbadness=\count268
+\c@finalcolumnbadness=\count269
+\last@try=\dimen147
+\multicolovershoot=\dimen148
+\multicolundershoot=\dimen149
+\mult@nat@firstbox=\box95
+\colbreak@box=\box96
+\mc@col@check@num=\count270
+)
+(/usr/share/texlive/texmf-dist/tex/latex/oberdiek/aliascnt.sty
+Package: aliascnt 2018/09/07 v1.5 Alias counters (HO)
+)
+\c@chapter=\count271
+LaTeX Font Info:    Redeclaring math symbol \Gamma on input line 379.
+LaTeX Font Info:    Redeclaring math symbol \Delta on input line 380.
+LaTeX Font Info:    Redeclaring math symbol \Theta on input line 381.
+LaTeX Font Info:    Redeclaring math symbol \Lambda on input line 382.
+LaTeX Font Info:    Redeclaring math symbol \Xi on input line 383.
+LaTeX Font Info:    Redeclaring math symbol \Pi on input line 384.
+LaTeX Font Info:    Redeclaring math symbol \Sigma on input line 385.
+LaTeX Font Info:    Redeclaring math symbol \Upsilon on input line 386.
+LaTeX Font Info:    Redeclaring math symbol \Phi on input line 387.
+LaTeX Font Info:    Redeclaring math symbol \Psi on input line 388.
+LaTeX Font Info:    Redeclaring math symbol \Omega on input line 389.
+LaTeX Info: Redefining \vec on input line 394.
+\tocchpnum=\dimen150
+\tocsecnum=\dimen151
+\tocsectotal=\dimen152
+\tocsubsecnum=\dimen153
+\tocsubsectotal=\dimen154
+\tocsubsubsecnum=\dimen155
+\tocsubsubsectotal=\dimen156
+\tocparanum=\dimen157
+\tocparatotal=\dimen158
+\tocsubparanum=\dimen159
+\@tempcntc=\count272
+\fnindent=\dimen160
+\c@@inst=\count273
+\c@@auth=\count274
+\c@auco=\count275
+\instindent=\dimen161
+\authrun=\box97
+\authorrunning=\toks17
+\tocauthor=\toks18
+\titrun=\box98
+\titlerunning=\toks19
+\toctitle=\toks20
+\c@theorem=\count276
+\c@case=\count277
+\c@conjecture=\count278
+\c@corollary=\count279
+\c@definition=\count280
+\c@example=\count281
+\c@exercise=\count282
+\c@lemma=\count283
+\c@note=\count284
+\c@problem=\count285
+\c@property=\count286
+\c@proposition=\count287
+\c@question=\count288
+\c@solution=\count289
+\c@remark=\count290
+\headlineindent=\dimen162
+)
+(/usr/share/texlive/texmf-dist/tex/latex/base/fontenc.sty
+Package: fontenc 2021/04/29 v2.0v Standard LaTeX package
+)
+(/usr/share/texlive/texmf-dist/tex/latex/graphics/graphicx.sty
+Package: graphicx 2021/09/16 v1.2d Enhanced LaTeX Graphics (DPC,SPQR)
+
+(/usr/share/texlive/texmf-dist/tex/latex/graphics/keyval.sty
+Package: keyval 2022/05/29 v1.15 key=value parser (DPC)
+\KV@toks@=\toks21
+)
+(/usr/share/texlive/texmf-dist/tex/latex/graphics/graphics.sty
+Package: graphics 2022/03/10 v1.4e Standard LaTeX Graphics (DPC,SPQR)
+
+(/usr/share/texlive/texmf-dist/tex/latex/graphics/trig.sty
+Package: trig 2021/08/11 v1.11 sin cos tan (DPC)
+)
+(/usr/share/texlive/texmf-dist/tex/latex/graphics-cfg/graphics.cfg
+File: graphics.cfg 2016/06/04 v1.11 sample graphics configuration
+)
+Package graphics Info: Driver file: pdftex.def on input line 107.
+
+(/usr/share/texlive/texmf-dist/tex/latex/graphics-def/pdftex.def
+File: pdftex.def 2022/09/22 v1.2b Graphics/color driver for pdftex
+))
+\Gin@req@height=\dimen163
+\Gin@req@width=\dimen164
+)
+(/usr/share/texlive/texmf-dist/tex/latex/booktabs/booktabs.sty
+Package: booktabs 2020/01/12 v1.61803398 Publication quality tables
+\heavyrulewidth=\dimen165
+\lightrulewidth=\dimen166
+\cmidrulewidth=\dimen167
+\belowrulesep=\dimen168
+\belowbottomsep=\dimen169
+\aboverulesep=\dimen170
+\abovetopsep=\dimen171
+\cmidrulesep=\dimen172
+\cmidrulekern=\dimen173
+\defaultaddspace=\dimen174
+\@cmidla=\count291
+\@cmidlb=\count292
+\@aboverulesep=\dimen175
+\@belowrulesep=\dimen176
+\@thisruleclass=\count293
+\@lastruleclass=\count294
+\@thisrulewidth=\dimen177
+)
+(/usr/share/texlive/texmf-dist/tex/latex/multirow/multirow.sty
+Package: multirow 2021/03/15 v2.8 Span multiple rows of a table
+\multirow@colwidth=\skip52
+\multirow@cntb=\count295
+\multirow@dima=\skip53
+\bigstrutjot=\dimen178
+)
+(/usr/share/texlive/texmf-dist/tex/latex/amsmath/amsmath.sty
+Package: amsmath 2023/05/13 v2.17o AMS math features
+\@mathmargin=\skip54
+
+For additional information on amsmath, use the `?' option.
+(/usr/share/texlive/texmf-dist/tex/latex/amsmath/amstext.sty
+Package: amstext 2021/08/26 v2.01 AMS text
+
+(/usr/share/texlive/texmf-dist/tex/latex/amsmath/amsgen.sty
+File: amsgen.sty 1999/11/30 v2.0 generic functions
+\@emptytoks=\toks22
+\ex@=\dimen179
+))
+(/usr/share/texlive/texmf-dist/tex/latex/amsmath/amsbsy.sty
+Package: amsbsy 1999/11/29 v1.2d Bold Symbols
+\pmbraise@=\dimen180
+)
+(/usr/share/texlive/texmf-dist/tex/latex/amsmath/amsopn.sty
+Package: amsopn 2022/04/08 v2.04 operator names
+)
+\inf@bad=\count296
+LaTeX Info: Redefining \frac on input line 234.
+\uproot@=\count297
+\leftroot@=\count298
+LaTeX Info: Redefining \overline on input line 399.
+LaTeX Info: Redefining \colon on input line 410.
+\classnum@=\count299
+\DOTSCASE@=\count300
+LaTeX Info: Redefining \ldots on input line 496.
+LaTeX Info: Redefining \dots on input line 499.
+LaTeX Info: Redefining \cdots on input line 620.
+\Mathstrutbox@=\box99
+\strutbox@=\box100
+LaTeX Info: Redefining \big on input line 722.
+LaTeX Info: Redefining \Big on input line 723.
+LaTeX Info: Redefining \bigg on input line 724.
+LaTeX Info: Redefining \Bigg on input line 725.
+\big@size=\dimen181
+LaTeX Font Info:    Redeclaring font encoding OML on input line 743.
+LaTeX Font Info:    Redeclaring font encoding OMS on input line 744.
+
+
+Package amsmath Warning: Unable to redefine math accent \vec.
+
+\macc@depth=\count301
+LaTeX Info: Redefining \bmod on input line 905.
+LaTeX Info: Redefining \pmod on input line 910.
+LaTeX Info: Redefining \smash on input line 940.
+LaTeX Info: Redefining \relbar on input line 970.
+LaTeX Info: Redefining \Relbar on input line 971.
+\c@MaxMatrixCols=\count302
+\dotsspace@=\muskip16
+\c@parentequation=\count303
+\dspbrk@lvl=\count304
+\tag@help=\toks23
+\row@=\count305
+\column@=\count306
+\maxfields@=\count307
+\andhelp@=\toks24
+\eqnshift@=\dimen182
+\alignsep@=\dimen183
+\tagshift@=\dimen184
+\tagwidth@=\dimen185
+\totwidth@=\dimen186
+\lineht@=\dimen187
+\@envbody=\toks25
+\multlinegap=\skip55
+\multlinetaggap=\skip56
+\mathdisplay@stack=\toks26
+LaTeX Info: Redefining \[ on input line 2953.
+LaTeX Info: Redefining \] on input line 2954.
+) (/usr/share/texlive/texmf-dist/tex/latex/url/url.sty
+\Urlmuskip=\muskip17
+Package: url 2013/09/16  ver 3.4  Verb mode for urls, etc.
+)
+(/usr/share/texlive/texmf-dist/tex/latex/adjustbox/adjustbox.sty
+Package: adjustbox 2022/10/17 v1.3a Adjusting TeX boxes (trim, clip, ...)
+
+(/usr/share/texlive/texmf-dist/tex/latex/xkeyval/xkeyval.sty
+Package: xkeyval 2022/06/16 v2.9 package option processing (HA)
+
+(/usr/share/texlive/texmf-dist/tex/generic/xkeyval/xkeyval.tex
+(/usr/share/texlive/texmf-dist/tex/generic/xkeyval/xkvutils.tex
+\XKV@toks=\toks27
+\XKV@tempa@toks=\toks28
+)
+\XKV@depth=\count308
+File: xkeyval.tex 2014/12/03 v2.7a key=value parser (HA)
+))
+(/usr/share/texlive/texmf-dist/tex/latex/adjustbox/adjcalc.sty
+Package: adjcalc 2012/05/16 v1.1 Provides advanced setlength with multiple back
+-ends (calc, etex, pgfmath)
+)
+(/usr/share/texlive/texmf-dist/tex/latex/adjustbox/trimclip.sty
+Package: trimclip 2020/08/19 v1.2 Trim and clip general TeX material
+
+(/usr/share/texlive/texmf-dist/tex/latex/collectbox/collectbox.sty
+Package: collectbox 2022/10/17 v0.4c Collect macro arguments as boxes
+\collectedbox=\box101
+)
+\tc@llx=\dimen188
+\tc@lly=\dimen189
+\tc@urx=\dimen190
+\tc@ury=\dimen191
+Package trimclip Info: Using driver 'tc-pdftex.def'.
+
+(/usr/share/texlive/texmf-dist/tex/latex/adjustbox/tc-pdftex.def
+File: tc-pdftex.def 2019/01/04 v2.2 Clipping driver for pdftex
+))
+\adjbox@Width=\dimen192
+\adjbox@Height=\dimen193
+\adjbox@Depth=\dimen194
+\adjbox@Totalheight=\dimen195
+\adjbox@pwidth=\dimen196
+\adjbox@pheight=\dimen197
+\adjbox@pdepth=\dimen198
+\adjbox@ptotalheight=\dimen199
+
+(/usr/share/texlive/texmf-dist/tex/latex/ifoddpage/ifoddpage.sty
+Package: ifoddpage 2022/10/18 v1.2 Conditionals for odd/even page detection
+\c@checkoddpage=\count309
+)
+(/usr/share/texlive/texmf-dist/tex/latex/varwidth/varwidth.sty
+Package: varwidth 2009/03/30 ver 0.92;  Variable-width minipages
+\@vwid@box=\box102
+\sift@deathcycles=\count310
+\@vwid@loff=\dimen256
+\@vwid@roff=\dimen257
+))
+(/usr/share/texlive/texmf-dist/tex/latex/pgf/frontendlayer/tikz.sty
+(/usr/share/texlive/texmf-dist/tex/latex/pgf/basiclayer/pgf.sty
+(/usr/share/texlive/texmf-dist/tex/latex/pgf/utilities/pgfrcs.sty
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/utilities/pgfutil-common.tex
+\pgfutil@everybye=\toks29
+\pgfutil@tempdima=\dimen258
+\pgfutil@tempdimb=\dimen259
+)
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/utilities/pgfutil-latex.def
+\pgfutil@abb=\box103
+)
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/utilities/pgfrcs.code.tex
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/pgf.revision.tex)
+Package: pgfrcs 2023-01-15 v3.1.10 (3.1.10)
+))
+Package: pgf 2023-01-15 v3.1.10 (3.1.10)
+
+(/usr/share/texlive/texmf-dist/tex/latex/pgf/basiclayer/pgfcore.sty
+(/usr/share/texlive/texmf-dist/tex/latex/pgf/systemlayer/pgfsys.sty
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/systemlayer/pgfsys.code.tex
+Package: pgfsys 2023-01-15 v3.1.10 (3.1.10)
+
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/utilities/pgfkeys.code.tex
+\pgfkeys@pathtoks=\toks30
+\pgfkeys@temptoks=\toks31
+
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/utilities/pgfkeyslibraryfiltered
+.code.tex
+\pgfkeys@tmptoks=\toks32
+))
+\pgf@x=\dimen260
+\pgf@y=\dimen261
+\pgf@xa=\dimen262
+\pgf@ya=\dimen263
+\pgf@xb=\dimen264
+\pgf@yb=\dimen265
+\pgf@xc=\dimen266
+\pgf@yc=\dimen267
+\pgf@xd=\dimen268
+\pgf@yd=\dimen269
+\w@pgf@writea=\write3
+\r@pgf@reada=\read2
+\c@pgf@counta=\count311
+\c@pgf@countb=\count312
+\c@pgf@countc=\count313
+\c@pgf@countd=\count314
+\t@pgf@toka=\toks33
+\t@pgf@tokb=\toks34
+\t@pgf@tokc=\toks35
+\pgf@sys@id@count=\count315
+ (/usr/share/texlive/texmf-dist/tex/generic/pgf/systemlayer/pgf.cfg
+File: pgf.cfg 2023-01-15 v3.1.10 (3.1.10)
+)
+Driver file for pgf: pgfsys-pdftex.def
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/systemlayer/pgfsys-pdftex.def
+File: pgfsys-pdftex.def 2023-01-15 v3.1.10 (3.1.10)
+
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/systemlayer/pgfsys-common-pdf.de
+f
+File: pgfsys-common-pdf.def 2023-01-15 v3.1.10 (3.1.10)
+)))
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/systemlayer/pgfsyssoftpath.code.
+tex
+File: pgfsyssoftpath.code.tex 2023-01-15 v3.1.10 (3.1.10)
+\pgfsyssoftpath@smallbuffer@items=\count316
+\pgfsyssoftpath@bigbuffer@items=\count317
+)
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/systemlayer/pgfsysprotocol.code.
+tex
+File: pgfsysprotocol.code.tex 2023-01-15 v3.1.10 (3.1.10)
+)) (/usr/share/texlive/texmf-dist/tex/latex/xcolor/xcolor.sty
+Package: xcolor 2023/11/15 v3.01 LaTeX color extensions (UK)
+
+(/usr/share/texlive/texmf-dist/tex/latex/graphics-cfg/color.cfg
+File: color.cfg 2016/01/02 v1.6 sample color configuration
+)
+Package xcolor Info: Driver file: pdftex.def on input line 274.
+
+(/usr/share/texlive/texmf-dist/tex/latex/graphics/mathcolor.ltx)
+Package xcolor Info: Model `cmy' substituted by `cmy0' on input line 1350.
+Package xcolor Info: Model `hsb' substituted by `rgb' on input line 1354.
+Package xcolor Info: Model `RGB' extended on input line 1366.
+Package xcolor Info: Model `HTML' substituted by `rgb' on input line 1368.
+Package xcolor Info: Model `Hsb' substituted by `hsb' on input line 1369.
+Package xcolor Info: Model `tHsb' substituted by `hsb' on input line 1370.
+Package xcolor Info: Model `HSB' substituted by `hsb' on input line 1371.
+Package xcolor Info: Model `Gray' substituted by `gray' on input line 1372.
+Package xcolor Info: Model `wave' substituted by `hsb' on input line 1373.
+)
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/basiclayer/pgfcore.code.tex
+Package: pgfcore 2023-01-15 v3.1.10 (3.1.10)
+
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/math/pgfmath.code.tex
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/math/pgfmathutil.code.tex)
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/math/pgfmathparser.code.tex
+\pgfmath@dimen=\dimen270
+\pgfmath@count=\count318
+\pgfmath@box=\box104
+\pgfmath@toks=\toks36
+\pgfmath@stack@operand=\toks37
+\pgfmath@stack@operation=\toks38
+)
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/math/pgfmathfunctions.code.tex)
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/math/pgfmathfunctions.basic.code
+.tex)
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/math/pgfmathfunctions.trigonomet
+ric.code.tex)
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/math/pgfmathfunctions.random.cod
+e.tex)
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/math/pgfmathfunctions.comparison
+.code.tex)
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/math/pgfmathfunctions.base.code.
+tex)
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/math/pgfmathfunctions.round.code
+.tex)
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/math/pgfmathfunctions.misc.code.
+tex)
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/math/pgfmathfunctions.integerari
+thmetics.code.tex)
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/math/pgfmathcalc.code.tex)
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/math/pgfmathfloat.code.tex
+\c@pgfmathroundto@lastzeros=\count319
+))
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/math/pgfint.code.tex)
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/basiclayer/pgfcorepoints.code.te
+x
+File: pgfcorepoints.code.tex 2023-01-15 v3.1.10 (3.1.10)
+\pgf@picminx=\dimen271
+\pgf@picmaxx=\dimen272
+\pgf@picminy=\dimen273
+\pgf@picmaxy=\dimen274
+\pgf@pathminx=\dimen275
+\pgf@pathmaxx=\dimen276
+\pgf@pathminy=\dimen277
+\pgf@pathmaxy=\dimen278
+\pgf@xx=\dimen279
+\pgf@xy=\dimen280
+\pgf@yx=\dimen281
+\pgf@yy=\dimen282
+\pgf@zx=\dimen283
+\pgf@zy=\dimen284
+)
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/basiclayer/pgfcorepathconstruct.
+code.tex
+File: pgfcorepathconstruct.code.tex 2023-01-15 v3.1.10 (3.1.10)
+\pgf@path@lastx=\dimen285
+\pgf@path@lasty=\dimen286
+)
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/basiclayer/pgfcorepathusage.code
+.tex
+File: pgfcorepathusage.code.tex 2023-01-15 v3.1.10 (3.1.10)
+\pgf@shorten@end@additional=\dimen287
+\pgf@shorten@start@additional=\dimen288
+)
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/basiclayer/pgfcorescopes.code.te
+x
+File: pgfcorescopes.code.tex 2023-01-15 v3.1.10 (3.1.10)
+\pgfpic=\box105
+\pgf@hbox=\box106
+\pgf@layerbox@main=\box107
+\pgf@picture@serial@count=\count320
+)
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/basiclayer/pgfcoregraphicstate.c
+ode.tex
+File: pgfcoregraphicstate.code.tex 2023-01-15 v3.1.10 (3.1.10)
+\pgflinewidth=\dimen289
+)
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/basiclayer/pgfcoretransformation
+s.code.tex
+File: pgfcoretransformations.code.tex 2023-01-15 v3.1.10 (3.1.10)
+\pgf@pt@x=\dimen290
+\pgf@pt@y=\dimen291
+\pgf@pt@temp=\dimen292
+)
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/basiclayer/pgfcorequick.code.tex
+File: pgfcorequick.code.tex 2023-01-15 v3.1.10 (3.1.10)
+)
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/basiclayer/pgfcoreobjects.code.t
+ex
+File: pgfcoreobjects.code.tex 2023-01-15 v3.1.10 (3.1.10)
+)
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/basiclayer/pgfcorepathprocessing
+.code.tex
+File: pgfcorepathprocessing.code.tex 2023-01-15 v3.1.10 (3.1.10)
+)
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/basiclayer/pgfcorearrows.code.te
+x
+File: pgfcorearrows.code.tex 2023-01-15 v3.1.10 (3.1.10)
+\pgfarrowsep=\dimen293
+)
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/basiclayer/pgfcoreshade.code.tex
+File: pgfcoreshade.code.tex 2023-01-15 v3.1.10 (3.1.10)
+\pgf@max=\dimen294
+\pgf@sys@shading@range@num=\count321
+\pgf@shadingcount=\count322
+)
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/basiclayer/pgfcoreimage.code.tex
+File: pgfcoreimage.code.tex 2023-01-15 v3.1.10 (3.1.10)
+)
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/basiclayer/pgfcoreexternal.code.
+tex
+File: pgfcoreexternal.code.tex 2023-01-15 v3.1.10 (3.1.10)
+\pgfexternal@startupbox=\box108
+)
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/basiclayer/pgfcorelayers.code.te
+x
+File: pgfcorelayers.code.tex 2023-01-15 v3.1.10 (3.1.10)
+)
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/basiclayer/pgfcoretransparency.c
+ode.tex
+File: pgfcoretransparency.code.tex 2023-01-15 v3.1.10 (3.1.10)
+)
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/basiclayer/pgfcorepatterns.code.
+tex
+File: pgfcorepatterns.code.tex 2023-01-15 v3.1.10 (3.1.10)
+)
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/basiclayer/pgfcorerdf.code.tex
+File: pgfcorerdf.code.tex 2023-01-15 v3.1.10 (3.1.10)
+)))
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/modules/pgfmoduleshapes.code.tex
+File: pgfmoduleshapes.code.tex 2023-01-15 v3.1.10 (3.1.10)
+\pgfnodeparttextbox=\box109
+) (/usr/share/texlive/texmf-dist/tex/generic/pgf/modules/pgfmoduleplot.code.tex
+File: pgfmoduleplot.code.tex 2023-01-15 v3.1.10 (3.1.10)
+)
+(/usr/share/texlive/texmf-dist/tex/latex/pgf/compatibility/pgfcomp-version-0-65
+.sty
+Package: pgfcomp-version-0-65 2023-01-15 v3.1.10 (3.1.10)
+\pgf@nodesepstart=\dimen295
+\pgf@nodesepend=\dimen296
+)
+(/usr/share/texlive/texmf-dist/tex/latex/pgf/compatibility/pgfcomp-version-1-18
+.sty
+Package: pgfcomp-version-1-18 2023-01-15 v3.1.10 (3.1.10)
+)) (/usr/share/texlive/texmf-dist/tex/latex/pgf/utilities/pgffor.sty
+(/usr/share/texlive/texmf-dist/tex/latex/pgf/utilities/pgfkeys.sty
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/utilities/pgfkeys.code.tex))
+(/usr/share/texlive/texmf-dist/tex/latex/pgf/math/pgfmath.sty
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/math/pgfmath.code.tex))
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/utilities/pgffor.code.tex
+Package: pgffor 2023-01-15 v3.1.10 (3.1.10)
+\pgffor@iter=\dimen297
+\pgffor@skip=\dimen298
+\pgffor@stack=\toks39
+\pgffor@toks=\toks40
+))
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/frontendlayer/tikz/tikz.code.tex
+Package: tikz 2023-01-15 v3.1.10 (3.1.10)
+
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/libraries/pgflibraryplothandlers
+.code.tex
+File: pgflibraryplothandlers.code.tex 2023-01-15 v3.1.10 (3.1.10)
+\pgf@plot@mark@count=\count323
+\pgfplotmarksize=\dimen299
+)
+\tikz@lastx=\dimen300
+\tikz@lasty=\dimen301
+\tikz@lastxsaved=\dimen302
+\tikz@lastysaved=\dimen303
+\tikz@lastmovetox=\dimen304
+\tikz@lastmovetoy=\dimen305
+\tikzleveldistance=\dimen306
+\tikzsiblingdistance=\dimen307
+\tikz@figbox=\box110
+\tikz@figbox@bg=\box111
+\tikz@tempbox=\box112
+\tikz@tempbox@bg=\box113
+\tikztreelevel=\count324
+\tikznumberofchildren=\count325
+\tikznumberofcurrentchild=\count326
+\tikz@fig@count=\count327
+
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/modules/pgfmodulematrix.code.tex
+File: pgfmodulematrix.code.tex 2023-01-15 v3.1.10 (3.1.10)
+\pgfmatrixcurrentrow=\count328
+\pgfmatrixcurrentcolumn=\count329
+\pgf@matrix@numberofcolumns=\count330
+)
+\tikz@expandcount=\count331
+
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/frontendlayer/tikz/libraries/tik
+zlibrarytopaths.code.tex
+File: tikzlibrarytopaths.code.tex 2023-01-15 v3.1.10 (3.1.10)
+)))
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/libraries/pgflibraryarrows.meta.
+code.tex
+File: pgflibraryarrows.meta.code.tex 2023-01-15 v3.1.10 (3.1.10)
+\pgfarrowinset=\dimen308
+\pgfarrowlength=\dimen309
+\pgfarrowwidth=\dimen310
+\pgfarrowlinewidth=\dimen311
+)
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/frontendlayer/tikz/libraries/tik
+zlibrarypositioning.code.tex
+File: tikzlibrarypositioning.code.tex 2023-01-15 v3.1.10 (3.1.10)
+)
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/frontendlayer/tikz/libraries/tik
+zlibraryshapes.geometric.code.tex
+File: tikzlibraryshapes.geometric.code.tex 2023-01-15 v3.1.10 (3.1.10)
+
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/libraries/shapes/pgflibraryshape
+s.geometric.code.tex
+File: pgflibraryshapes.geometric.code.tex 2023-01-15 v3.1.10 (3.1.10)
+))
+(/usr/share/texlive/texmf-dist/tex/generic/pgf/frontendlayer/tikz/libraries/tik
+zlibraryfit.code.tex
+File: tikzlibraryfit.code.tex 2023-01-15 v3.1.10 (3.1.10)
+)
+(/usr/share/texlive/texmf-dist/tex/latex/l3backend/l3backend-pdftex.def
+File: l3backend-pdftex.def 2024-01-04 L3 backend support: PDF output (pdfTeX)
+\l__color_backend_stack_int=\count332
+\l__pdf_internal_box=\box114
+)
+(./main.aux)
+\openout1 = `main.aux'.
+
+LaTeX Font Info:    Checking defaults for OML/cmm/m/it on input line 21.
+LaTeX Font Info:    ... okay on input line 21.
+LaTeX Font Info:    Checking defaults for OMS/cmsy/m/n on input line 21.
+LaTeX Font Info:    ... okay on input line 21.
+LaTeX Font Info:    Checking defaults for OT1/cmr/m/n on input line 21.
+LaTeX Font Info:    ... okay on input line 21.
+LaTeX Font Info:    Checking defaults for T1/cmr/m/n on input line 21.
+LaTeX Font Info:    ... okay on input line 21.
+LaTeX Font Info:    Checking defaults for TS1/cmr/m/n on input line 21.
+LaTeX Font Info:    ... okay on input line 21.
+LaTeX Font Info:    Checking defaults for OMX/cmex/m/n on input line 21.
+LaTeX Font Info:    ... okay on input line 21.
+LaTeX Font Info:    Checking defaults for U/cmr/m/n on input line 21.
+LaTeX Font Info:    ... okay on input line 21.
+ (/usr/share/texlive/texmf-dist/tex/context/base/mkii/supp-pdf.mkii
+[Loading MPS to PDF converter (version 2006.09.02).]
+\scratchcounter=\count333
+\scratchdimen=\dimen312
+\scratchbox=\box115
+\nofMPsegments=\count334
+\nofMParguments=\count335
+\everyMPshowfont=\toks41
+\MPscratchCnt=\count336
+\MPscratchDim=\dimen313
+\MPnumerator=\count337
+\makeMPintoPDFobject=\count338
+\everyMPtoPDFconversion=\toks42
+) (/usr/share/texlive/texmf-dist/tex/latex/epstopdf-pkg/epstopdf-base.sty
+Package: epstopdf-base 2020-01-24 v2.11 Base part for package epstopdf
+Package epstopdf-base Info: Redefining graphics rule for `.eps' on input line 4
+85.
+
+(/usr/share/texlive/texmf-dist/tex/latex/latexconfig/epstopdf-sys.cfg
+File: epstopdf-sys.cfg 2010/07/13 v1.3 Configuration of (r)epstopdf for TeX Liv
+e
+))
+[1
+
+{/var/lib/texmf/fonts/map/pdftex/updmap/pdftex.map}{/usr/share/texmf/fonts/enc/
+dvips/cm-super/cm-super-t1.enc}{/usr/share/texmf/fonts/enc/dvips/cm-super/cm-su
+per-ts1.enc}] [2] [3] [4] [5] [6] [7]
+LaTeX Font Info:    Trying to load font information for T1+cmtt on input line 3
+02.
+
+(/usr/share/texlive/texmf-dist/tex/latex/base/t1cmtt.fd
+File: t1cmtt.fd 2023/04/13 v2.5m Standard LaTeX font definitions
+) [8] (./main.aux)
+ ***********
+LaTeX2e <2023-11-01> patch level 1
+L3 programming layer <2024-01-22>
+ ***********
+ ) 
+Here is how much of TeX's memory you used:
+ 15625 strings out of 474222
+ 314614 string characters out of 5748734
+ 1945975 words of memory out of 5000000
+ 37598 multiletter control sequences out of 15000+600000
+ 568379 words of font info for 57 fonts, out of 8000000 for 9000
+ 1141 hyphenation exceptions out of 8191
+ 84i,11n,93p,1573b,809s stack positions out of 10000i,1000n,20000p,200000b,200000s
+</usr/share/texlive/texmf-dist/fonts/type1/public/amsfonts/cm/cmmi10.pfb></us
+r/share/texlive/texmf-dist/fonts/type1/public/amsfonts/cm/cmmi7.pfb></usr/share
+/texlive/texmf-dist/fonts/type1/public/amsfonts/cm/cmr10.pfb></usr/share/texliv
+e/texmf-dist/fonts/type1/public/amsfonts/cm/cmr7.pfb></usr/share/texlive/texmf-
+dist/fonts/type1/public/amsfonts/cm/cmsy10.pfb></usr/share/texlive/texmf-dist/f
+onts/type1/public/amsfonts/cm/cmsy7.pfb></usr/share/texlive/texmf-dist/fonts/ty
+pe1/public/amsfonts/cm/cmsy9.pfb></usr/share/texmf/fonts/type1/public/cm-super/
+sfbx0900.pfb></usr/share/texmf/fonts/type1/public/cm-super/sfbx1000.pfb></usr/s
+hare/texmf/fonts/type1/public/cm-super/sfbx1200.pfb></usr/share/texmf/fonts/typ
+e1/public/cm-super/sfbx1440.pfb></usr/share/texmf/fonts/type1/public/cm-super/s
+frm0700.pfb></usr/share/texmf/fonts/type1/public/cm-super/sfrm0900.pfb></usr/sh
+are/texmf/fonts/type1/public/cm-super/sfrm1000.pfb></usr/share/texmf/fonts/type
+1/public/cm-super/sftt0900.pfb>
+Output written on main.pdf (8 pages, 282223 bytes).
+PDF statistics:
+ 114 PDF objects out of 1000 (max. 8388607)
+ 72 compressed objects within 1 object stream
+ 0 named destinations out of 1000 (max. 500000)
+ 13 words of extra memory for PDF output out of 10000 (max. 10000000)
+

paper/main.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d208b41809dc236697cea51fc0aa93b8e941d7211e869c4813277ef53acfce97
+size 282223

paper/main.tex

@@ -0,0 +1,309 @@
+\documentclass[runningheads]{llncs}
+
+\usepackage[T1]{fontenc}
+\usepackage{graphicx}
+\usepackage{booktabs}
+\usepackage{multirow}
+\usepackage{amsmath}
+\usepackage{url}
+\usepackage{adjustbox}
+\usepackage{tikz}
+\usetikzlibrary{arrows.meta,positioning,shapes.geometric,fit}
+\emergencystretch=2em
+
+\title{DynaFall: Robust Skeleton-Based Fall Detection via Keypoint Dynamics and Graph-Based Benchmarking}
+\titlerunning{DynaFall for Skeleton-Based Fall Detection}
+
+\author{Anonymous Submission}
+\authorrunning{Anonymous}
+\institute{}
+
+\begin{document}
+\maketitle
+
+\begin{abstract}
+Automatic fall detection is a safety-critical problem for elderly-care and assisted-living environments, but vision-based systems must balance accuracy, privacy, robustness, and computational cost. This paper presents DynaFall-GCN, a skeleton-based fall-detection framework that converts RGB videos into 17-keypoint pose sequences and performs classification only on normalized skeleton features. Unlike generic action-recognition models that mainly learn joint trajectories, DynaFall-GCN explicitly combines three complementary streams: joint coordinates, bone vectors, and fall-oriented dynamics including velocity, acceleration, torso angle, hip-drop motion, body aspect ratio, and center-of-body displacement. A confidence-aware keypoint dropout strategy is further used to expose the model to pose-estimation failures during training. To make the evaluation stronger than a single-model report, we benchmark DynaFall-GCN against LSTM, ST-GCN, 2s-AGCN-style adaptive GCN, CTR-GCN-style refined GCN, PoseC3D-style heatmap CNN, and a TCN--Transformer baseline on URFD and MCFD. Experiments are conducted over three random seeds with video-level URFD splits and scenario-level MCFD splits. DynaFall-GCN achieves the best URFD F1 and macro-F1, and the highest MCFD fall recall among the evaluated methods. Ablation and efficiency results show that explicit dynamics are the most reliable contributor to clean-set performance, while confidence-aware dropout shifts the detector toward higher sensitivity under challenging multi-view conditions.
+\keywords{Fall detection \and Skeleton action recognition \and Graph convolutional networks \and Keypoint dynamics \and Pose estimation}
+\end{abstract}
+
+\section{Introduction}
+Falls are a major risk for older adults because delayed assistance after a fall can lead to severe injury, long-lie complications, loss of independence, and higher care cost. Vision-based fall detection is attractive because it does not require the user to wear or recharge a device, and recent surveys show that deep learning has become a dominant direction for non-intrusive fall monitoring~\cite{alam2022review}. However, raw RGB video also introduces practical concerns: appearance cues may encode private information, background and clothing can bias a classifier, and large video models can be expensive for continuous deployment.
+
+Skeleton-based recognition is a pragmatic compromise. A pose estimator first converts each frame into body keypoints, after which the fall classifier only sees motion geometry rather than raw appearance. This representation is compact, privacy-friendlier, and naturally aligned with the physical structure of a fall: rapid center-of-body motion, torso rotation, hip descent, and a transition from upright to horizontal posture. At the same time, skeleton-based fall detection is not solved by simply applying a small recurrent model. Pose keypoints can be noisy near the floor, cameras can observe the same fall from very different viewpoints, and fall datasets are usually small and imbalanced. These issues motivate a careful benchmark, not only a proposed model.
+
+This work focuses on two established fall datasets: UR Fall Detection (URFD)~\cite{urfd} and Multiple Cameras Fall Dataset (MCFD)~\cite{mcfd}. URFD is useful for controlled RGB-to-skeleton experiments, while MCFD stresses multi-view generalization. The complete processing pipeline is shown in Fig.~\ref{fig:pipeline}. We use a consistent protocol: all available videos are converted to pose clips, train/validation/test splits are formed before clip generation, and all compared methods use the same clips, optimizer, weighted focal loss, and seeds.
+
+The contributions are threefold. First, we build a unified RGB-to-skeleton benchmark for URFD and MCFD with a leakage-aware split protocol. Second, we compare representative sequence, graph, heatmap, and temporal-attention baselines under the same experimental setting. Third, we propose DynaFall-GCN, a lightweight three-stream model that augments joint and bone features with fall-specific dynamics and confidence-aware keypoint dropout. Tables~\ref{tab:datasets}--\ref{tab:ablation} and Fig.~\ref{fig:runtime} report the dataset statistics, main results, ablation, and efficiency trade-offs.
+
+\begin{figure}[t]
+\centering
+\begin{tikzpicture}[
+  node distance=0.55cm,
+  block/.style={draw, rounded corners, align=center, minimum height=0.72cm, minimum width=2.15cm, font=\scriptsize},
+  arrow/.style={-{Latex[length=2mm]}, thick}
+]
+\node[block] (rgb) {RGB video};
+\node[block, right=of rgb] (sample) {Frame sampling\\$T=32$};
+\node[block, right=of sample] (pose) {YOLO pose\\17 keypoints};
+\node[block, right=of pose] (norm) {Box-centered\\normalization};
+\node[block, below=of norm] (feat) {Joint + bone\\+ dynamics};
+\node[block, left=of feat] (models) {Baselines and\\DynaFall-GCN};
+\node[block, left=of models] (pred) {Fall /\\non-fall};
+\draw[arrow] (rgb) -- (sample);
+\draw[arrow] (sample) -- (pose);
+\draw[arrow] (pose) -- (norm);
+\draw[arrow] (norm) -- (feat);
+\draw[arrow] (feat) -- (models);
+\draw[arrow] (models) -- (pred);
+\end{tikzpicture}
+\caption{Overall RGB-to-skeleton fall-detection pipeline used for all experiments.}
+\label{fig:pipeline}
+\end{figure}
+
+\section{Related Work}
+\textbf{Vision-based fall detection.}
+Deep vision-based fall detection has moved from handcrafted motion and silhouette descriptors to CNN, recurrent, temporal-convolution, transformer, and pose-based models. A recent review emphasizes that non-intrusive vision systems are promising but still face dataset bias, class imbalance, viewpoint variation, occlusion, and real-time deployment constraints~\cite{alam2022review}. These issues motivate our evaluation protocol: Table~\ref{tab:main} reports accuracy together with precision, recall, fall-class F1, and macro-F1 instead of accuracy alone.
+
+\textbf{Fall datasets and evaluation protocol.}
+URFD combines falls and activities of daily living with RGB, depth, and inertial measurements~\cite{urfd}. MCFD contains multi-camera fall scenarios and confounding non-fall activities observed from eight cameras~\cite{mcfd}. Because these datasets are small, leakage prevention is critical. In MCFD, camera-level splitting can place different views of the same scenario into both training and testing, so we use scenario-level splitting as summarized in Table~\ref{tab:datasets}.
+
+\textbf{Graph-based skeleton recognition.}
+Skeleton action recognition provides stronger building blocks than classical fall-detection baselines. ST-GCN models joints and bones as a spatial-temporal graph~\cite{yan2018stgcn}. 2s-AGCN improves this by learning adaptive topology and adding a bone stream~\cite{shi2019agcn}, while CTR-GCN refines topology channel-wise for more flexible joint aggregation~\cite{chen2021ctrgcn}. These models motivate our joint and bone streams, but they are designed for general action recognition rather than fall-specific motion mechanics.
+
+\textbf{Heatmap and temporal-attention alternatives.}
+PoseC3D/PoseConv3D converts skeletons into 3D heatmap volumes and applies 3D CNNs, giving a non-GCN comparison for pose-based recognition~\cite{duan2022posec3d}. More recently, TCNTE combines temporal convolutions and Transformer Encoder modules for real-time skeleton-based fall detection and uses weighted focal loss for imbalance~\cite{yu2025tcnte}. Our TCN--Transformer baseline follows this motivation, while DynaFall-GCN adds explicit fall-oriented measurements such as torso angle and hip-drop motion.
+
+\textbf{Position of this work.}
+The main gap is that strong skeleton action-recognition models and fall-specific temporal models are rarely compared under one leakage-aware RGB-to-skeleton protocol. In addition, many methods rely on learned temporal features but do not explicitly encode fall mechanics. DynaFall-GCN is designed to fill this gap: it keeps the compactness of skeleton models, borrows the joint/bone reasoning that makes graph models effective, and adds dynamics features that are physically tied to falls. Fig.~\ref{fig:architecture} summarizes this design.
+
+\section{Method}
+\subsection{Pose Extraction and Normalization}
+Each RGB frame is processed by YOLOv8 pose~\cite{yolov8}. If multiple people are detected, the person with the largest bounding box is selected because both URFD and MCFD focus on one dominant subject. Each frame is represented by 17 COCO keypoints with $(x,y,c)$ coordinates, where $c$ is the pose confidence. Coordinates are normalized by the visible person box:
+\begin{equation}
+x' = \frac{x - x_c}{w}, \qquad y' = \frac{y - y_c}{h},
+\end{equation}
+where $(x_c,y_c)$ is the box center and $(w,h)$ is the box size. Videos are converted into clips of 32 frames with stride 16; shorter sequences are padded by repeating the final frame.
+
+\subsection{Feature Construction}
+For a clip with normalized joints $P_t \in \mathrm{R}^{17 \times 3}$ at time $t$, DynaFall-GCN builds three complementary feature groups. The joint feature is the normalized keypoint tensor itself. The bone feature is computed from the COCO skeleton edges:
+\begin{equation}
+B_t^{(i,j)} = P_t^{j} - P_t^{i},
+\end{equation}
+where $(i,j)$ is a connected joint pair. This representation emphasizes body configuration, limb direction, and posture change. The dynamics feature is designed around fall mechanics. It includes first-order velocity $V_t=P_t-P_{t-1}$, second-order acceleration $A_t=V_t-V_{t-1}$, torso angle from shoulder and hip centers, hip vertical displacement, body aspect ratio, and center-of-body velocity. These descriptors encode the rapid downward movement and upright-to-horizontal transition that distinguish a fall from many activities of daily living.
+
+\subsection{DynaFall-GCN}
+DynaFall-GCN uses three streams as shown in Fig.~\ref{fig:architecture}. The joint stream receives normalized keypoints and learns spatial-temporal dependencies over the human-body graph. The bone stream receives bone vectors and focuses on posture and limb orientation. The dynamics stream receives the fall-oriented features above and models short-term motion using temporal convolutions followed by attention pooling. Joint and bone streams use lightweight graph-temporal encoders with adaptive adjacency so that the model can learn both physical skeleton links and task-dependent joint relations. The resulting embeddings are concatenated and classified by a multilayer perceptron.
+
+\begin{figure}[t]
+\centering
+\begin{tikzpicture}[
+  node distance=0.38cm and 0.7cm,
+  block/.style={draw, rounded corners, align=center, minimum height=0.64cm, minimum width=2.25cm, font=\scriptsize},
+  stream/.style={draw, rounded corners, align=center, minimum height=0.64cm, minimum width=2.55cm, font=\scriptsize},
+  arrow/.style={-{Latex[length=2mm]}, thick}
+]
+\node[block] (input) {Pose clip\\$32 \times 17 \times 3$};
+\node[block, below=of input] (drop) {Confidence-aware\\keypoint dropout};
+\node[stream, below left=0.6cm and 1.8cm of drop] (joint) {Joint GCN\\coordinates};
+\node[stream, below=0.6cm of drop] (bone) {Bone GCN\\bone vectors};
+\node[stream, below right=0.6cm and 1.8cm of drop] (dyn) {Dynamics TCN\\fall cues};
+\node[block, below=1.05cm of bone] (fusion) {Feature fusion\\concatenation};
+\node[block, below=of fusion] (head) {MLP classifier};
+\node[block, below=of head] (out) {Fall / non-fall};
+\draw[arrow] (input) -- (drop);
+\draw[arrow] (drop) -- (joint);
+\draw[arrow] (drop) -- (bone);
+\draw[arrow] (drop) -- (dyn);
+\draw[arrow] (joint) -- (fusion);
+\draw[arrow] (bone) -- (fusion);
+\draw[arrow] (dyn) -- (fusion);
+\draw[arrow] (fusion) -- (head);
+\draw[arrow] (head) -- (out);
+\end{tikzpicture}
+\caption{DynaFall-GCN architecture with joint, bone, and fall-specific dynamics streams.}
+\label{fig:architecture}
+\end{figure}
+
+\subsection{Confidence-Aware Keypoint Dropout}
+Pose estimation can fail during occlusion, motion blur, or when the body is close to the floor. During training, DynaFall applies confidence-aware keypoint dropout: high-confidence joints are masked with probability 0.100, while joints with confidence below 0.300 are masked with probability 0.500. Masking sets coordinates, confidence, and derived motion values to zero. Weighted focal loss~\cite{lin2017focal} is used for all methods to reduce class-imbalance effects.
+
+\section{Experimental Setup}
+\subsection{Datasets, Splits, and Metrics}
+URFD is evaluated with video-level splits. MCFD is evaluated with scenario-level splits so that camera views of one scenario never appear in different splits. Table~\ref{tab:datasets} summarizes the processed RGB-to-pose data used in our experiments. We use seeds 7, 13, and 21. For each seed, the split ratio is 70\% train, 15\% validation, and 15\% test at the video or scenario level. Unless otherwise stated, precision, recall, and F1-score are computed for the positive fall class:
+\begin{equation}
+F1_{\mathrm{fall}}=\frac{2P_{\mathrm{fall}}R_{\mathrm{fall}}}{P_{\mathrm{fall}}+R_{\mathrm{fall}}}.
+\end{equation}
+Macro-F1 is the unweighted mean of the fall-class and non-fall-class F1 scores. Recall is reported explicitly because missing a true fall is usually more costly than producing a false alarm.
+
+\begin{table}[t]
+\centering
+\caption{Dataset statistics after RGB video collection and pose extraction.}
+\label{tab:datasets}
+\small
+\begin{adjustbox}{width=\textwidth}
+\begin{tabular}{lrrrrrl}
+\toprule
+Dataset & Scenarios & Videos & Fall videos & Non-fall videos & Frames & Split unit \\
+\midrule
+URFD & -- & 70 & 30 & 40 & 11,936 & Video \\
+MCFD & 24 & 192 & 176 & 16 & 261,139 & Scenario \\
+\bottomrule
+\end{tabular}
+\end{adjustbox}
+\end{table}
+
+\subsection{Compared Methods}
+All models use the same extracted pose clips, optimizer, loss, and seed protocol. The baselines are intentionally diverse: LSTM tests a simple sequence model, ST-GCN tests fixed graph reasoning, 2s-AGCN and CTR-GCN test stronger adaptive graph reasoning, PoseC3D tests a heatmap representation, and TCNTE tests a recent fall-oriented temporal design. DynaFall-GCN is evaluated as the proposed joint--bone--dynamics model.
+
+\section{Results}
+\subsection{Main Comparison}
+Table~\ref{tab:main} reports mean and standard deviation across three seeds. On URFD, DynaFall-GCN obtains the best accuracy, fall-class F1, and macro-F1. This indicates that explicit dynamics are useful when pose quality is relatively stable and the clips are short. On MCFD, the scenario-level split is harder because the same physical fall is not shared across cameras between train and test. Under this stricter setting, AGCN gives the highest accuracy and macro-F1, while DynaFall-GCN reaches the highest fall recall and F1. This trade-off is important: DynaFall-GCN is more sensitive to falls, whereas AGCN is more balanced across both classes.
+
+\begin{table}[t]
+\centering
+\caption{Main comparison over three seeds. Precision, recall, and F1 are computed for the fall class.}
+\label{tab:main}
+\small
+\begin{adjustbox}{width=\textwidth}
+\begin{tabular}{llccccc}
+\toprule
+Dataset & Method & Acc. & Prec. & Recall & F1 & Macro-F1 \\
+\midrule
+\multirow{7}{*}{URFD}
+& LSTM & 0.762$\pm$0.051 & 0.519$\pm$0.131 & 0.654$\pm$0.196 & 0.565$\pm$0.118 & 0.700$\pm$0.069 \\
+& ST-GCN & 0.742$\pm$0.059 & 0.480$\pm$0.126 & 0.614$\pm$0.167 & 0.531$\pm$0.130 & 0.676$\pm$0.081 \\
+& AGCN & 0.652$\pm$0.206 & 0.467$\pm$0.223 & \textbf{0.813$\pm$0.162} & 0.556$\pm$0.143 & 0.626$\pm$0.185 \\
+& CTR-GCN & 0.732$\pm$0.138 & \textbf{0.587$\pm$0.292} & 0.696$\pm$0.233 & 0.568$\pm$0.045 & 0.679$\pm$0.096 \\
+& PoseC3D & 0.626$\pm$0.083 & 0.320$\pm$0.079 & 0.560$\pm$0.393 & 0.386$\pm$0.164 & 0.545$\pm$0.040 \\
+& TCNTE & 0.702$\pm$0.169 & 0.489$\pm$0.265 & 0.718$\pm$0.156 & 0.558$\pm$0.203 & 0.665$\pm$0.170 \\
+& DynaFall & \textbf{0.770$\pm$0.032} & 0.524$\pm$0.012 & 0.714$\pm$0.179 & \textbf{0.597$\pm$0.053} & \textbf{0.716$\pm$0.014} \\
+\midrule
+\multirow{7}{*}{MCFD}
+& LSTM & 0.643$\pm$0.002 & 0.588$\pm$0.047 & 0.567$\pm$0.123 & 0.571$\pm$0.065 & 0.630$\pm$0.018 \\
+& ST-GCN & 0.644$\pm$0.015 & 0.601$\pm$0.030 & 0.500$\pm$0.011 & 0.546$\pm$0.019 & 0.626$\pm$0.005 \\
+& AGCN & \textbf{0.670$\pm$0.005} & \textbf{0.611$\pm$0.057} & 0.642$\pm$0.025 & 0.625$\pm$0.017 & \textbf{0.665$\pm$0.007} \\
+& CTR-GCN & 0.587$\pm$0.047 & 0.510$\pm$0.054 & 0.768$\pm$0.077 & 0.613$\pm$0.062 & 0.581$\pm$0.043 \\
+& PoseC3D & 0.553$\pm$0.086 & 0.522$\pm$0.145 & 0.583$\pm$0.270 & 0.509$\pm$0.118 & 0.538$\pm$0.077 \\
+& TCNTE & 0.633$\pm$0.019 & 0.564$\pm$0.038 & 0.645$\pm$0.038 & 0.600$\pm$0.020 & 0.630$\pm$0.018 \\
+& DynaFall & 0.616$\pm$0.057 & 0.541$\pm$0.059 & \textbf{0.772$\pm$0.115} & \textbf{0.631$\pm$0.048} & 0.607$\pm$0.060 \\
+\bottomrule
+\end{tabular}
+\end{adjustbox}
+\end{table}
+
+\subsection{Ablation Study}
+Table~\ref{tab:ablation} studies DynaFall-GCN components. Joint-only modeling is a reasonable baseline, but adding bone information improves URFD performance because bone vectors capture posture and body orientation. Adding explicit dynamics gives the most consistent clean-set gains: it improves F1 and macro-F1 relative to the joint-only and joint+bone variants on both datasets. Random dropout mainly increases recall on URFD, while confidence-aware dropout strongly increases MCFD recall but can reduce macro-F1. This supports a deployment-oriented interpretation: when missed falls are unacceptable, sensitivity-oriented training may be preferred, but threshold calibration is needed to control false alarms.
+
+\begin{table}[t]
+\centering
+\caption{DynaFall-GCN ablation over three seeds. Precision, recall, and F1 are fall-class metrics.}
+\label{tab:ablation}
+\small
+\begin{adjustbox}{width=\textwidth}
+\begin{tabular}{llccccc}
+\toprule
+Dataset & Variant & Acc. & Prec. & Recall & F1 & Macro-F1 \\
+\midrule
+\multirow{5}{*}{URFD}
+& A1 Joint & 0.666$\pm$0.040 & 0.394$\pm$0.050 & 0.731$\pm$0.226 & 0.507$\pm$0.098 & 0.623$\pm$0.030 \\
+& A2 Joint+Bone & 0.748$\pm$0.115 & 0.563$\pm$0.192 & 0.658$\pm$0.253 & 0.558$\pm$0.118 & 0.687$\pm$0.098 \\
+& A3 +Dynamics & \textbf{0.769$\pm$0.095} & \textbf{0.618$\pm$0.337} & 0.614$\pm$0.063 & \textbf{0.576$\pm$0.118} & \textbf{0.708$\pm$0.095} \\
+& A4 +Random Dropout & 0.727$\pm$0.087 & 0.507$\pm$0.112 & \textbf{0.752$\pm$0.297} & 0.565$\pm$0.059 & 0.675$\pm$0.020 \\
+& A5 +Confidence Dropout & 0.663$\pm$0.128 & 0.411$\pm$0.077 & 0.703$\pm$0.274 & 0.503$\pm$0.085 & 0.611$\pm$0.092 \\
+\midrule
+\multirow{5}{*}{MCFD}
+& A1 Joint & 0.631$\pm$0.008 & 0.581$\pm$0.053 & 0.517$\pm$0.083 & 0.543$\pm$0.040 & 0.617$\pm$0.015 \\
+& A2 Joint+Bone & 0.630$\pm$0.075 & 0.589$\pm$0.074 & 0.567$\pm$0.161 & 0.562$\pm$0.071 & 0.611$\pm$0.062 \\
+& A3 +Dynamics & \textbf{0.649$\pm$0.013} & \textbf{0.591$\pm$0.045} & 0.605$\pm$0.061 & 0.595$\pm$0.016 & \textbf{0.642$\pm$0.007} \\
+& A4 +Random Dropout & 0.605$\pm$0.025 & 0.536$\pm$0.033 & 0.566$\pm$0.059 & 0.550$\pm$0.034 & 0.598$\pm$0.025 \\
+& A5 +Confidence Dropout & 0.597$\pm$0.106 & 0.537$\pm$0.118 & \textbf{0.708$\pm$0.030} & \textbf{0.604$\pm$0.071} & 0.594$\pm$0.109 \\
+\bottomrule
+\end{tabular}
+\end{adjustbox}
+\end{table}
+
+\subsection{Runtime and Efficiency}
+Fig.~\ref{fig:runtime} replaces the runtime table with a visual efficiency summary. All methods are lightweight compared with RGB video models because they operate on 32-frame skeleton clips. PoseC3D has the smallest parameter count in this compact implementation, but it also gives the lowest mean F1. DynaFall-GCN has 0.307M parameters, remains above 1000 clips/s on URFD and 4200 clips/s on MCFD, and achieves the best average fall-class F1 across the two datasets. This makes it a practical compromise between accuracy and efficiency.
+
+\begin{figure}[t]
+\centering
+\resizebox{\textwidth}{!}{%
+\begin{tikzpicture}[x=0.68cm,y=3.6cm]
+\scriptsize
+\draw[->] (-0.45,0) -- (6.85,0);
+\draw[->] (-0.45,0) -- (-0.45,0.66);
+\foreach \y/\lab in {0.0/0.000,0.2/0.200,0.4/0.400,0.6/0.600}
+  \draw (-0.50,\y) -- (-0.40,\y) node[left=1pt] {\lab};
+\foreach \name/\x/\f in {LSTM/0/0.568,ST-GCN/1/0.539,AGCN/2/0.590,CTR-GCN/3/0.590,PoseC3D/4/0.447,TCNTE/5/0.579,DynaFall/6/0.614} {
+  \draw[fill=black!55] (\x-0.24,0) rectangle (\x+0.24,\f);
+  \node[rotate=35, anchor=east] at (\x,-0.035) {\name};
+}
+\node[rotate=90] at (-1.05,0.33) {Mean fall-class F1};
+\begin{scope}[xshift=6.8cm,x=0.68cm,y=0.82cm]
+\draw[->] (-0.45,0) -- (6.85,0);
+\draw[->] (-0.45,0) -- (-0.45,5.0);
+\foreach \y/\lab in {1/1k,2/2k,3/3k,4/4k}
+  \draw (-0.50,\y) -- (-0.40,\y) node[left=1pt] {\lab};
+\foreach \name/\x/\u/\m in {LSTM/0/1.235/4.268,ST-GCN/1/1.126/4.249,AGCN/2/1.081/4.363,CTR-GCN/3/1.060/4.231,PoseC3D/4/1.023/3.978,TCNTE/5/1.086/4.334,DynaFall/6/1.024/4.285} {
+  \draw[fill=black!25] (\x-0.22,0) rectangle (\x-0.03,\u);
+  \draw[fill=black!65] (\x+0.03,0) rectangle (\x+0.22,\m);
+  \node[rotate=35, anchor=east] at (\x,-0.24) {\name};
+}
+\node[fill=black!25, minimum width=0.30cm, minimum height=0.12cm] at (4.45,4.80) {};
+\node[right] at (4.65,4.80) {URFD};
+\node[fill=black!65, minimum width=0.30cm, minimum height=0.12cm] at (5.45,4.80) {};
+\node[right] at (5.65,4.80) {MCFD};
+\node[rotate=90] at (-1.05,2.5) {clips/s};
+\end{scope}
+\end{tikzpicture}
+}
+\caption{Runtime and efficiency summary. Left: mean fall-class F1 across URFD and MCFD. Right: inference throughput in clips/s.}
+\label{fig:runtime}
+\end{figure}
+
+\section{Discussion}
+The results show two operating regimes. On URFD, DynaFall-GCN performs best because pose quality is stable and explicit posture-transition cues are reliable. On MCFD, scenario-level splitting makes the problem harder: test clips come from unseen scenarios and viewpoints. AGCN therefore gives stronger macro-F1, while DynaFall-GCN still gives the highest fall recall. This is useful in safety monitoring, where a sensitivity-oriented model can be preferable if decision thresholds are calibrated to control false alarms.
+
+The ablation in Table~\ref{tab:ablation} explains the model behavior. Bone features encode body configuration, while dynamics features encode fast downward motion and loss of upright support. Confidence-aware dropout is more nuanced: it does not always improve clean-set macro-F1, but it raises MCFD recall, making it suitable when missed-fall reduction is the priority.
+
+\section{Conclusion}
+This paper presented DynaFall-GCN, a skeleton-based fall detector that combines joint, bone, and fall-specific dynamics features with confidence-aware keypoint dropout. Under a unified RGB-to-skeleton protocol on URFD and MCFD, DynaFall-GCN achieves the best URFD fall-class F1 and macro-F1 and the highest MCFD fall recall, while remaining lightweight. Future work should test real non-staged falls, improve cross-view calibration, and integrate on-device pose estimation.
+
+\begingroup
+\footnotesize
+\begin{thebibliography}{10}
+\setlength{\itemsep}{0pt}
+\setlength{\parsep}{0pt}
+\bibitem{alam2022review}
+Alam, E., Sufian, A., Dutta, P., Leo, M.: Vision-based human fall detection systems using deep learning: A review. Comput. Biol. Med. \textbf{146}, 105626 (2022)
+
+\bibitem{urfd}
+Kepski, M., Kwolek, B.: Human fall detection on embedded platform using depth maps and wireless accelerometer. Comput. Methods Programs Biomed. \textbf{117}(3), 489--501 (2014)
+
+\bibitem{mcfd}
+Auvinet, E., Multon, F., Saint-Arnaud, A., Rousseau, J., Meunier, J.: Fall detection using body geometry and human motion analysis. In: ICADI (2011)
+
+\bibitem{yan2018stgcn}
+Yan, S., Xiong, Y., Lin, D.: Spatial temporal graph convolutional networks for skeleton-based action recognition. In: AAAI (2018)
+
+\bibitem{shi2019agcn}
+Shi, L., Zhang, Y., Cheng, J., Lu, H.: Two-stream adaptive graph convolutional networks for skeleton-based action recognition. In: CVPR (2019)
+
+\bibitem{chen2021ctrgcn}
+Chen, Y., Zhang, Z., Yuan, C., Li, B., Deng, Y., Hu, W.: Channel-wise topology refinement graph convolution for skeleton-based action recognition. In: ICCV (2021)
+
+\bibitem{duan2022posec3d}
+Duan, H., Zhao, Y., Chen, K., Shao, D., Lin, D., Dai, B.: Revisiting skeleton-based action recognition. In: CVPR (2022)
+
+\bibitem{yu2025tcnte}
+Yu, X., Wang, C., Wu, W., Xiong, S.: Real-time skeleton-based fall detection using temporal convolutional networks and transformer encoder. Pervasive Mob. Comput. \textbf{102}, 102016 (2025)
+
+\bibitem{yolov8}
+Ultralytics: YOLOv8. \url{https://github.com/ultralytics/ultralytics} (2023)
+
+\bibitem{lin2017focal}
+Lin, T.Y., Goyal, P., Girshick, R., He, K., Dollar, P.: Focal loss for dense object detection. In: ICCV, pp. 2980--2988 (2017)
+\end{thebibliography}
+\endgroup
+
+\end{document}

paper/references.bib

@@ -0,0 +1,77 @@
+@article{alam2022review,
+  title={Vision-Based Human Fall Detection Systems Using Deep Learning: A Review},
+  author={Alam, Ekram and Sufian, Abu and Dutta, Paramartha and Leo, Marco},
+  journal={Computers in Biology and Medicine},
+  volume={146},
+  pages={105626},
+  year={2022}
+}
+
+@article{urfd,
+  title={Human Fall Detection on Embedded Platform Using Depth Maps and Wireless Accelerometer},
+  author={Kepski, Michal and Kwolek, Bogdan},
+  journal={Computer Methods and Programs in Biomedicine},
+  volume={117},
+  number={3},
+  pages={489--501},
+  year={2014}
+}
+
+@inproceedings{mcfd,
+  title={Fall Detection Using Body Geometry and Human Motion Analysis},
+  author={Auvinet, Edouard and Multon, Franck and Saint-Arnaud, Alain and Rousseau, Jacqueline and Meunier, Jean},
+  booktitle={International Conference on Aging, Disability and Independence},
+  year={2011}
+}
+
+@inproceedings{yan2018stgcn,
+  title={Spatial Temporal Graph Convolutional Networks for Skeleton-Based Action Recognition},
+  author={Yan, Sijie and Xiong, Yuanjun and Lin, Dahua},
+  booktitle={AAAI Conference on Artificial Intelligence},
+  year={2018}
+}
+
+@inproceedings{shi2019agcn,
+  title={Two-Stream Adaptive Graph Convolutional Networks for Skeleton-Based Action Recognition},
+  author={Shi, Lei and Zhang, Yifan and Cheng, Jian and Lu, Hanqing},
+  booktitle={IEEE/CVF Conference on Computer Vision and Pattern Recognition},
+  year={2019}
+}
+
+@inproceedings{chen2021ctrgcn,
+  title={Channel-Wise Topology Refinement Graph Convolution for Skeleton-Based Action Recognition},
+  author={Chen, Yuxin and Zhang, Ziqi and Yuan, Chunfeng and Li, Bing and Deng, Ying and Hu, Weiming},
+  booktitle={IEEE/CVF International Conference on Computer Vision},
+  year={2021}
+}
+
+@inproceedings{duan2022posec3d,
+  title={Revisiting Skeleton-Based Action Recognition},
+  author={Duan, Haodong and Zhao, Yue and Chen, Kai and Shao, Dahua and Lin, Dahua and Dai, Bo},
+  booktitle={IEEE/CVF Conference on Computer Vision and Pattern Recognition},
+  year={2022}
+}
+
+@article{yu2025tcnte,
+  title={A Real-Time Skeleton-Based Fall Detection Algorithm Based on Temporal Convolutional Networks and Transformer Encoder},
+  author={Yu, Xiaoqun and Wang, Chenfeng and Wu, Wenyu and Xiong, Shuping},
+  journal={Pervasive and Mobile Computing},
+  volume={102},
+  pages={102016},
+  year={2025}
+}
+
+@misc{yolov8,
+  title={Ultralytics YOLOv8},
+  author={{Ultralytics}},
+  year={2023},
+  howpublished={\url{https://github.com/ultralytics/ultralytics}}
+}
+
+@inproceedings{lin2017focal,
+  title={Focal Loss for Dense Object Detection},
+  author={Lin, Tsung-Yi and Goyal, Priya and Girshick, Ross and He, Kaiming and Dollar, Piotr},
+  booktitle={IEEE International Conference on Computer Vision},
+  pages={2980--2988},
+  year={2017}
+}

requirements.txt

@@ -0,0 +1,8 @@
+numpy>=2.0.0
+pandas>=2.2.0
+scikit-learn>=1.5.0
+torch>=2.4.0
+opencv-python>=4.10.0
+ultralytics>=8.3.0
+PyYAML>=6.0.0
+tqdm>=4.66.0

results/MCFD/agcn/best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b543ce3bce34b29f41c2528d6a2fd0d95eb6c527b391bc759fad4e23eb68c94e
+size 867809

results/MCFD/agcn/history.json

@@ -0,0 +1,178 @@
+[
+  {
+    "epoch": 1,
+    "loss": 0.10004046732046974,
+    "val_fps": 3134.4307448275636,
+    "val_accuracy": 0.3697478991596639,
+    "val_precision": 0.9356060606060606,
+    "val_recall": 0.15632911392405063,
+    "val_specificity": 0.9697508896797153,
+    "val_f1": 0.2678958785249458,
+    "val_macro_f1": 0.3573085950001778
+  },
+  {
+    "epoch": 2,
+    "loss": 0.0952134209069903,
+    "val_fps": 3989.2144291239,
+    "val_accuracy": 0.438375350140056,
+    "val_precision": 0.9586374695863747,
+    "val_recall": 0.24936708860759493,
+    "val_specificity": 0.9697508896797153,
+    "val_f1": 0.39578101456554493,
+    "val_macro_f1": 0.43557040261639657
+  },
+  {
+    "epoch": 3,
+    "loss": 0.09230573286970654,
+    "val_fps": 4118.985734226191,
+    "val_accuracy": 0.44677871148459386,
+    "val_precision": 0.9540229885057471,
+    "val_recall": 0.2626582278481013,
+    "val_specificity": 0.9644128113879004,
+    "val_f1": 0.4119106699751861,
+    "val_macro_f1": 0.44482708465705095
+  },
+  {
+    "epoch": 4,
+    "loss": 0.08984891978091163,
+    "val_fps": 3756.089846434585,
+    "val_accuracy": 0.5088702147525677,
+    "val_precision": 0.9012158054711246,
+    "val_recall": 0.37531645569620253,
+    "val_specificity": 0.8843416370106761,
+    "val_f1": 0.5299374441465594,
+    "val_macro_f1": 0.5078817230507968
+  },
+  {
+    "epoch": 5,
+    "loss": 0.08813056671797974,
+    "val_fps": 3946.4217176482607,
+    "val_accuracy": 0.5004668534080299,
+    "val_precision": 0.909967845659164,
+    "val_recall": 0.35822784810126584,
+    "val_specificity": 0.900355871886121,
+    "val_f1": 0.5140781108083561,
+    "val_macro_f1": 0.5000745981515363
+  },
+  {
+    "epoch": 6,
+    "loss": 0.08850927808761767,
+    "val_fps": 3977.7542180338896,
+    "val_accuracy": 0.6718020541549953,
+    "val_precision": 0.8968325791855204,
+    "val_recall": 0.6272151898734177,
+    "val_specificity": 0.797153024911032,
+    "val_f1": 0.7381750465549348,
+    "val_macro_f1": 0.6492626327208695
+  },
+  {
+    "epoch": 7,
+    "loss": 0.0875586135081745,
+    "val_fps": 4063.5771939174692,
+    "val_accuracy": 0.41876750700280113,
+    "val_precision": 0.9970326409495549,
+    "val_recall": 0.21265822784810126,
+    "val_specificity": 0.998220640569395,
+    "val_f1": 0.350547730829421,
+    "val_macro_f1": 0.4122827374045711
+  },
+  {
+    "epoch": 8,
+    "loss": 0.08657941715353396,
+    "val_fps": 4129.31866378651,
+    "val_accuracy": 0.45518207282913165,
+    "val_precision": 0.9619686800894854,
+    "val_recall": 0.2721518987341772,
+    "val_specificity": 0.9697508896797153,
+    "val_f1": 0.4242723236309818,
+    "val_macro_f1": 0.4536071409913881
+  },
+  {
+    "epoch": 9,
+    "loss": 0.08473612372889669,
+    "val_fps": 4016.697393823691,
+    "val_accuracy": 0.450046685340803,
+    "val_precision": 0.9808612440191388,
+    "val_recall": 0.25949367088607594,
+    "val_specificity": 0.9857651245551602,
+    "val_f1": 0.41041041041041043,
+    "val_macro_f1": 0.44754991211684125
+  },
+  {
+    "epoch": 10,
+    "loss": 0.08441147508469016,
+    "val_fps": 4224.448727227852,
+    "val_accuracy": 0.41876750700280113,
+    "val_precision": 0.9827089337175793,
+    "val_recall": 0.2158227848101266,
+    "val_specificity": 0.9893238434163701,
+    "val_f1": 0.353918007265179,
+    "val_macro_f1": 0.41285208806194884
+  },
+  {
+    "epoch": 11,
+    "loss": 0.08371498034509714,
+    "val_fps": 3946.110962148607,
+    "val_accuracy": 0.45471521942110177,
+    "val_precision": 0.9598214285714286,
+    "val_recall": 0.2721518987341772,
+    "val_specificity": 0.9679715302491103,
+    "val_f1": 0.4240631163708087,
+    "val_macro_f1": 0.45316630995845397
+  },
+  {
+    "epoch": 12,
+    "loss": 0.08121818716447805,
+    "val_fps": 4093.604896901611,
+    "val_accuracy": 0.5028011204481793,
+    "val_precision": 0.9432013769363167,
+    "val_recall": 0.3468354430379747,
+    "val_specificity": 0.9412811387900356,
+    "val_f1": 0.5071726052753355,
+    "val_macro_f1": 0.5027619974092175
+  },
+  {
+    "epoch": 13,
+    "loss": 0.08042114470340046,
+    "val_fps": 4088.9173874767603,
+    "val_accuracy": 0.465452847805789,
+    "val_precision": 0.9718004338394793,
+    "val_recall": 0.28354430379746837,
+    "val_specificity": 0.9768683274021353,
+    "val_f1": 0.43900048995590396,
+    "val_macro_f1": 0.4642617251384513
+  },
+  {
+    "epoch": 14,
+    "loss": 0.07871276757197611,
+    "val_fps": 3936.6529261234623,
+    "val_accuracy": 0.511204481792717,
+    "val_precision": 0.9333333333333333,
+    "val_recall": 0.3632911392405063,
+    "val_specificity": 0.9270462633451957,
+    "val_f1": 0.5230068337129841,
+    "val_macro_f1": 0.5109050444294935
+  },
+  {
+    "epoch": 15,
+    "loss": 0.07788032612706217,
+    "val_fps": 4058.6236299958105,
+    "val_accuracy": 0.5634920634920635,
+    "val_precision": 0.9271523178807947,
+    "val_recall": 0.4430379746835443,
+    "val_specificity": 0.902135231316726,
+    "val_f1": 0.5995717344753747,
+    "val_macro_f1": 0.5599192689821717
+  },
+  {
+    "epoch": 16,
+    "loss": 0.0775182641240185,
+    "val_fps": 4044.822823388296,
+    "val_accuracy": 0.569094304388422,
+    "val_precision": 0.9121706398996235,
+    "val_recall": 0.460126582278481,
+    "val_specificity": 0.8754448398576512,
+    "val_f1": 0.6116954143878839,
+    "val_macro_f1": 0.563844560890848
+  }
+]

results/MCFD/agcn/metrics_test_clean.json

@@ -0,0 +1,14 @@
+{
+  "dataset": "MCFD",
+  "method": "agcn",
+  "split": "test",
+  "robustness": "clean",
+  "fps": 4337.702795788146,
+  "params": 208520,
+  "accuracy": 0.6852678571428571,
+  "precision": 0.8030412744388125,
+  "recall": 0.6589423648247178,
+  "specificity": 0.7293532338308458,
+  "f1": 0.7238903394255874,
+  "macro_f1": 0.6789866922041432
+}

results/MCFD/ctrgcn/best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:685e6073cd104bf13b391c5eac183652a5a683bbbfe68ca62f61e5258da61f40
+size 1535165

results/MCFD/ctrgcn/history.json

@@ -0,0 +1,178 @@
+[
+  {
+    "epoch": 1,
+    "loss": 0.10115521957037622,
+    "val_fps": 3916.4185443643028,
+    "val_accuracy": 0.5560224089635855,
+    "val_precision": 0.8887515451174289,
+    "val_recall": 0.4550632911392405,
+    "val_specificity": 0.8398576512455516,
+    "val_f1": 0.6019254918375889,
+    "val_macro_f1": 0.5500392630691903
+  },
+  {
+    "epoch": 2,
+    "loss": 0.0979851155612251,
+    "val_fps": 3930.0279331313577,
+    "val_accuracy": 0.3412698412698413,
+    "val_precision": 0.9941520467836257,
+    "val_recall": 0.10759493670886076,
+    "val_specificity": 0.998220640569395,
+    "val_f1": 0.1941747572815534,
+    "val_macro_f1": 0.3185638887078908
+  },
+  {
+    "epoch": 3,
+    "loss": 0.09610443587246192,
+    "val_fps": 3983.2557743756643,
+    "val_accuracy": 0.4696545284780579,
+    "val_precision": 0.8639344262295082,
+    "val_recall": 0.33354430379746836,
+    "val_specificity": 0.8523131672597865,
+    "val_f1": 0.4812785388127854,
+    "val_macro_f1": 0.46938807551909567
+  },
+  {
+    "epoch": 4,
+    "loss": 0.09342363508107318,
+    "val_fps": 4143.628063231366,
+    "val_accuracy": 0.39822595704948643,
+    "val_precision": 0.9833887043189369,
+    "val_recall": 0.18734177215189873,
+    "val_specificity": 0.9911032028469751,
+    "val_f1": 0.3147262094630516,
+    "val_macro_f1": 0.38915669607567893
+  },
+  {
+    "epoch": 5,
+    "loss": 0.09133928454083826,
+    "val_fps": 4006.6160123772993,
+    "val_accuracy": 0.4084967320261438,
+    "val_precision": 0.9264305177111717,
+    "val_recall": 0.21518987341772153,
+    "val_specificity": 0.9519572953736655,
+    "val_f1": 0.3492552645095018,
+    "val_macro_f1": 0.4035536057250119
+  },
+  {
+    "epoch": 6,
+    "loss": 0.09130121225998677,
+    "val_fps": 3958.759546833165,
+    "val_accuracy": 0.6241830065359477,
+    "val_precision": 0.8918099089989889,
+    "val_recall": 0.5582278481012658,
+    "val_specificity": 0.8096085409252669,
+    "val_f1": 0.6866485013623979,
+    "val_macro_f1": 0.6086303731301785
+  },
+  {
+    "epoch": 7,
+    "loss": 0.08848687835813589,
+    "val_fps": 3879.0833979112163,
+    "val_accuracy": 0.5859010270774977,
+    "val_precision": 0.8811881188118812,
+    "val_recall": 0.5069620253164557,
+    "val_specificity": 0.8078291814946619,
+    "val_f1": 0.6436319807151466,
+    "val_macro_f1": 0.5747407814439243
+  },
+  {
+    "epoch": 8,
+    "loss": 0.08724830102457129,
+    "val_fps": 3976.8811610127977,
+    "val_accuracy": 0.45704948646125115,
+    "val_precision": 0.9464668094218416,
+    "val_recall": 0.27974683544303797,
+    "val_specificity": 0.9555160142348754,
+    "val_f1": 0.4318514899853444,
+    "val_macro_f1": 0.4559793882649118
+  },
+  {
+    "epoch": 9,
+    "loss": 0.08591482387098368,
+    "val_fps": 3958.81335050971,
+    "val_accuracy": 0.4393090569561158,
+    "val_precision": 0.9501187648456056,
+    "val_recall": 0.25316455696202533,
+    "val_specificity": 0.9626334519572953,
+    "val_f1": 0.39980009995002497,
+    "val_macro_f1": 0.43686895054443864
+  },
+  {
+    "epoch": 10,
+    "loss": 0.08534977833252831,
+    "val_fps": 3910.2815384837972,
+    "val_accuracy": 0.42810457516339867,
+    "val_precision": 0.947103274559194,
+    "val_recall": 0.2379746835443038,
+    "val_specificity": 0.9626334519572953,
+    "val_f1": 0.38037430450177034,
+    "val_macro_f1": 0.4246908366895501
+  },
+  {
+    "epoch": 11,
+    "loss": 0.08508303862453018,
+    "val_fps": 3933.3533502889377,
+    "val_accuracy": 0.4098972922502334,
+    "val_precision": 0.898989898989899,
+    "val_recall": 0.22531645569620254,
+    "val_specificity": 0.9288256227758007,
+    "val_f1": 0.3603238866396761,
+    "val_macro_f1": 0.4063317873406353
+  },
+  {
+    "epoch": 12,
+    "loss": 0.0829382371077694,
+    "val_fps": 4062.3340534008466,
+    "val_accuracy": 0.5219421101774043,
+    "val_precision": 0.9186746987951807,
+    "val_recall": 0.3860759493670886,
+    "val_specificity": 0.9039145907473309,
+    "val_f1": 0.5436720142602496,
+    "val_macro_f1": 0.520855614973262
+  },
+  {
+    "epoch": 13,
+    "loss": 0.08236354132968077,
+    "val_fps": 3984.950983377803,
+    "val_accuracy": 0.4169000933706816,
+    "val_precision": 0.9635854341736695,
+    "val_recall": 0.21772151898734177,
+    "val_specificity": 0.9768683274021353,
+    "val_f1": 0.3551884357253485,
+    "val_macro_f1": 0.4115098548460573
+  },
+  {
+    "epoch": 14,
+    "loss": 0.08053340387136042,
+    "val_fps": 4090.831176365699,
+    "val_accuracy": 0.49533146591970123,
+    "val_precision": 0.9528130671506352,
+    "val_recall": 0.3322784810126582,
+    "val_specificity": 0.9537366548042705,
+    "val_f1": 0.4927264195213515,
+    "val_macro_f1": 0.4953181563468346
+  },
+  {
+    "epoch": 15,
+    "loss": 0.07969231103758159,
+    "val_fps": 3818.7140201431444,
+    "val_accuracy": 0.6190476190476191,
+    "val_precision": 0.9116379310344828,
+    "val_recall": 0.5354430379746835,
+    "val_specificity": 0.8540925266903915,
+    "val_f1": 0.6746411483253588,
+    "val_macro_f1": 0.6075908444329496
+  },
+  {
+    "epoch": 16,
+    "loss": 0.07879969071113264,
+    "val_fps": 3918.3975140927073,
+    "val_accuracy": 0.5084033613445378,
+    "val_precision": 0.9428571428571428,
+    "val_recall": 0.3550632911392405,
+    "val_specificity": 0.9395017793594306,
+    "val_f1": 0.5158620689655172,
+    "val_macro_f1": 0.50828665325943
+  }
+]

results/MCFD/ctrgcn/metrics_test_clean.json

@@ -0,0 +1,14 @@
+{
+  "dataset": "MCFD",
+  "method": "ctrgcn",
+  "split": "test",
+  "robustness": "clean",
+  "fps": 4142.889289836818,
+  "params": 374984,
+  "accuracy": 0.6655505952380952,
+  "precision": 0.8091482649842271,
+  "recall": 0.6096256684491979,
+  "specificity": 0.7592039800995025,
+  "f1": 0.6953575059301932,
+  "macro_f1": 0.6623179282228285
+}

results/MCFD/lstm/best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b352091be2adaf0c60542a81d08f801d4e0d1bb5fa0486a4a6514151d9897b1f
+size 462662

results/MCFD/lstm/history.json

@@ -0,0 +1,35 @@
+[
+  {
+    "epoch": 1,
+    "loss": 0.13412656985786603,
+    "val_fps": 3971.9944165583393,
+    "val_accuracy": 0.7703081232492998,
+    "val_precision": 0.7706467661691542,
+    "val_recall": 0.980379746835443,
+    "val_specificity": 0.1797153024911032,
+    "val_f1": 0.8629526462395544,
+    "val_macro_f1": 0.5770094643301518
+  },
+  {
+    "epoch": 2,
+    "loss": 0.12558948784301352,
+    "val_fps": 3955.109527686606,
+    "val_accuracy": 0.7754435107376284,
+    "val_precision": 0.7988036976617727,
+    "val_recall": 0.9297468354430379,
+    "val_specificity": 0.3416370106761566,
+    "val_f1": 0.8593155893536122,
+    "val_macro_f1": 0.6516231125958812
+  },
+  {
+    "epoch": 3,
+    "loss": 0.11973153530339711,
+    "val_fps": 4123.892664527366,
+    "val_accuracy": 0.7670401493930906,
+    "val_precision": 0.8144269924374636,
+    "val_recall": 0.8860759493670886,
+    "val_specificity": 0.43238434163701067,
+    "val_f1": 0.8487420430433464,
+    "val_macro_f1": 0.6710715291358864
+  }
+]

results/MCFD/lstm/metrics_test_clean.json

@@ -0,0 +1,14 @@
+{
+  "dataset": "MCFD",
+  "method": "lstm",
+  "split": "test",
+  "robustness": "clean",
+  "fps": 4186.835521132716,
+  "params": 114818,
+  "accuracy": 0.6517857142857143,
+  "precision": 0.6488640892785971,
+  "recall": 0.9673202614379085,
+  "specificity": 0.12338308457711443,
+  "f1": 0.7767175572519084,
+  "macro_f1": 0.4930885083556839
+}

results/MCFD/posec3d/best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b3d035ec1bc13904954aff1829e814c475b3f84cda93848a7ef7f58f67619ff4
+size 63921

results/MCFD/posec3d/history.json

@@ -0,0 +1,244 @@
+[
+  {
+    "epoch": 1,
+    "loss": 0.10217119398900174,
+    "val_fps": 3656.50104612873,
+    "val_accuracy": 0.28291316526610644,
+    "val_precision": 0.84375,
+    "val_recall": 0.03417721518987342,
+    "val_specificity": 0.9822064056939501,
+    "val_f1": 0.06569343065693431,
+    "val_macro_f1": 0.24193762441937622
+  },
+  {
+    "epoch": 2,
+    "loss": 0.09731143623186586,
+    "val_fps": 3547.91571222601,
+    "val_accuracy": 0.26283846872082167,
+    "val_precision": 1.0,
+    "val_recall": 0.0006329113924050633,
+    "val_specificity": 1.0,
+    "val_f1": 0.001265022137887413,
+    "val_macro_f1": 0.20854964018474098
+  },
+  {
+    "epoch": 3,
+    "loss": 0.09620521088979724,
+    "val_fps": 3642.8764576204126,
+    "val_accuracy": 0.32866479925303455,
+    "val_precision": 0.9493670886075949,
+    "val_recall": 0.0949367088607595,
+    "val_specificity": 0.9857651245551602,
+    "val_f1": 0.17261219792865362,
+    "val_macro_f1": 0.30390232834374553
+  },
+  {
+    "epoch": 4,
+    "loss": 0.09444303353570838,
+    "val_fps": 3610.701546726241,
+    "val_accuracy": 0.3141923436041083,
+    "val_precision": 1.0,
+    "val_recall": 0.07025316455696203,
+    "val_specificity": 1.0,
+    "val_f1": 0.13128326434062684,
+    "val_macro_f1": 0.2823790020121954
+  },
+  {
+    "epoch": 5,
+    "loss": 0.09451987360007583,
+    "val_fps": 3783.594217081561,
+    "val_accuracy": 0.5070028011204482,
+    "val_precision": 0.8540540540540541,
+    "val_recall": 0.4,
+    "val_specificity": 0.8078291814946619,
+    "val_f1": 0.5448275862068965,
+    "val_macro_f1": 0.5035746892337944
+  },
+  {
+    "epoch": 6,
+    "loss": 0.09331116356625196,
+    "val_fps": 3657.283943711066,
+    "val_accuracy": 0.30438842203548083,
+    "val_precision": 1.0,
+    "val_recall": 0.056962025316455694,
+    "val_specificity": 1.0,
+    "val_f1": 0.10778443113772455,
+    "val_macro_f1": 0.2688883900141568
+  },
+  {
+    "epoch": 7,
+    "loss": 0.09331376635909251,
+    "val_fps": 3639.6938368377128,
+    "val_accuracy": 0.2880485527544351,
+    "val_precision": 1.0,
+    "val_recall": 0.03481012658227848,
+    "val_specificity": 1.0,
+    "val_f1": 0.0672782874617737,
+    "val_macro_f1": 0.24579467411971284
+  },
+  {
+    "epoch": 8,
+    "loss": 0.09261355322709948,
+    "val_fps": 3646.997401587179,
+    "val_accuracy": 0.3520074696545285,
+    "val_precision": 0.9571428571428572,
+    "val_recall": 0.12721518987341773,
+    "val_specificity": 0.9839857651245552,
+    "val_f1": 0.2245810055865922,
+    "val_macro_f1": 0.33402265997052144
+  },
+  {
+    "epoch": 9,
+    "loss": 0.09247916392163612,
+    "val_fps": 3784.38539006663,
+    "val_accuracy": 0.46078431372549017,
+    "val_precision": 0.841091492776886,
+    "val_recall": 0.33164556962025316,
+    "val_specificity": 0.8238434163701067,
+    "val_f1": 0.47571493418066274,
+    "val_macro_f1": 0.46034665498076865
+  },
+  {
+    "epoch": 10,
+    "loss": 0.09164904068608087,
+    "val_fps": 3746.254735996759,
+    "val_accuracy": 0.44864612511671337,
+    "val_precision": 0.9235668789808917,
+    "val_recall": 0.27531645569620256,
+    "val_specificity": 0.9359430604982206,
+    "val_f1": 0.42418332520721597,
+    "val_macro_f1": 0.44764920850598144
+  },
+  {
+    "epoch": 11,
+    "loss": 0.09135045006488597,
+    "val_fps": 3730.431618928083,
+    "val_accuracy": 0.28057889822595705,
+    "val_precision": 1.0,
+    "val_recall": 0.024683544303797468,
+    "val_specificity": 1.0,
+    "val_f1": 0.04817788758492897,
+    "val_macro_f1": 0.23497074491816805
+  },
+  {
+    "epoch": 12,
+    "loss": 0.09147095468198181,
+    "val_fps": 3683.6886127132557,
+    "val_accuracy": 0.5859010270774977,
+    "val_precision": 0.7194426852438252,
+    "val_recall": 0.7189873417721518,
+    "val_specificity": 0.2117437722419929,
+    "val_f1": 0.7192149414371637,
+    "val_macro_f1": 0.46538524849635965
+  },
+  {
+    "epoch": 13,
+    "loss": 0.09101124636074276,
+    "val_fps": 3724.331666140878,
+    "val_accuracy": 0.33286647992530344,
+    "val_precision": 1.0,
+    "val_recall": 0.09556962025316455,
+    "val_specificity": 1.0,
+    "val_f1": 0.17446562680531486,
+    "val_macro_f1": 0.3073659900575732
+  },
+  {
+    "epoch": 14,
+    "loss": 0.09165158919915663,
+    "val_fps": 3787.654097807469,
+    "val_accuracy": 0.4780578898225957,
+    "val_precision": 0.8357558139534884,
+    "val_recall": 0.3639240506329114,
+    "val_specificity": 0.798932384341637,
+    "val_f1": 0.5070546737213404,
+    "val_macro_f1": 0.47624559082892415
+  },
+  {
+    "epoch": 15,
+    "loss": 0.09028075870323113,
+    "val_fps": 3679.6292521915902,
+    "val_accuracy": 0.3618113912231559,
+    "val_precision": 0.9819004524886877,
+    "val_recall": 0.13734177215189874,
+    "val_specificity": 0.9928825622775801,
+    "val_f1": 0.24097723486951694,
+    "val_macro_f1": 0.3452167688644806
+  },
+  {
+    "epoch": 16,
+    "loss": 0.09013592564472969,
+    "val_fps": 3802.481168526258,
+    "val_accuracy": 0.35527544351073764,
+    "val_precision": 0.9853658536585366,
+    "val_recall": 0.12784810126582277,
+    "val_specificity": 0.994661921708185,
+    "val_f1": 0.22633053221288515,
+    "val_macro_f1": 0.3368547418967587
+  },
+  {
+    "epoch": 17,
+    "loss": 0.09048822282490648,
+    "val_fps": 3644.9348088242136,
+    "val_accuracy": 0.3104575163398693,
+    "val_precision": 1.0,
+    "val_recall": 0.06518987341772152,
+    "val_specificity": 1.0,
+    "val_f1": 0.12240047534165181,
+    "val_macro_f1": 0.27727097969312503
+  },
+  {
+    "epoch": 18,
+    "loss": 0.089875658604711,
+    "val_fps": 3727.8665597599506,
+    "val_accuracy": 0.32866479925303455,
+    "val_precision": 1.0,
+    "val_recall": 0.08987341772151898,
+    "val_specificity": 1.0,
+    "val_f1": 0.16492450638792103,
+    "val_macro_f1": 0.30182212829154054
+  },
+  {
+    "epoch": 19,
+    "loss": 0.08958346086863785,
+    "val_fps": 3658.1168641926465,
+    "val_accuracy": 0.5760971055088702,
+    "val_precision": 0.7094763092269327,
+    "val_recall": 0.720253164556962,
+    "val_specificity": 0.1708185053380783,
+    "val_f1": 0.714824120603015,
+    "val_macro_f1": 0.4446847875742348
+  },
+  {
+    "epoch": 20,
+    "loss": 0.08948933731727356,
+    "val_fps": 3824.6645835849695,
+    "val_accuracy": 0.373015873015873,
+    "val_precision": 0.9876543209876543,
+    "val_recall": 0.1518987341772152,
+    "val_specificity": 0.994661921708185,
+    "val_f1": 0.2633022490400439,
+    "val_macro_f1": 0.3587945621470029
+  },
+  {
+    "epoch": 21,
+    "loss": 0.08952507244345975,
+    "val_fps": 3620.5778052596556,
+    "val_accuracy": 0.4976657329598506,
+    "val_precision": 0.794392523364486,
+    "val_recall": 0.43037974683544306,
+    "val_specificity": 0.6868327402135231,
+    "val_f1": 0.5582922824302134,
+    "val_macro_f1": 0.4880206000895656
+  },
+  {
+    "epoch": 22,
+    "loss": 0.08912930570638469,
+    "val_fps": 3608.4476960904117,
+    "val_accuracy": 0.3828197945845005,
+    "val_precision": 0.9961538461538462,
+    "val_recall": 0.1639240506329114,
+    "val_specificity": 0.998220640569395,
+    "val_f1": 0.28152173913043477,
+    "val_macro_f1": 0.3703026044261012
+  }
+]

results/MCFD/posec3d/metrics_test_clean.json

@@ -0,0 +1,14 @@
+{
+  "dataset": "MCFD",
+  "method": "posec3d",
+  "split": "test",
+  "robustness": "clean",
+  "fps": 3741.17066541071,
+  "params": 14466,
+  "accuracy": 0.5070684523809523,
+  "precision": 0.5920781893004116,
+  "recall": 0.6838978015448604,
+  "specificity": 0.2109452736318408,
+  "f1": 0.6346843121036669,
+  "macro_f1": 0.4385542772639547
+}

results/MCFD/stgcn/best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ed20268699b7d9bb897cb94594c908f2be6fd9f1587ddcab93bbcd837cfad559
+size 430405

results/MCFD/stgcn/history.json

@@ -0,0 +1,35 @@
+[
+  {
+    "epoch": 1,
+    "loss": 0.1345855339766399,
+    "val_fps": 3749.472548126596,
+    "val_accuracy": 0.7226890756302521,
+    "val_precision": 0.7879672897196262,
+    "val_recall": 0.8537974683544304,
+    "val_specificity": 0.3540925266903915,
+    "val_f1": 0.8195625759416768,
+    "val_macro_f1": 0.6103861266805158
+  },
+  {
+    "epoch": 2,
+    "loss": 0.12828448044930646,
+    "val_fps": 3952.838195876346,
+    "val_accuracy": 0.7100840336134454,
+    "val_precision": 0.8565055762081785,
+    "val_recall": 0.7291139240506329,
+    "val_specificity": 0.6565836298932385,
+    "val_f1": 0.7876923076923077,
+    "val_macro_f1": 0.6653693326541008
+  },
+  {
+    "epoch": 3,
+    "loss": 0.12449156771946736,
+    "val_fps": 3674.3579758648907,
+    "val_accuracy": 0.6844070961718021,
+    "val_precision": 0.8662884927066451,
+    "val_recall": 0.6765822784810127,
+    "val_specificity": 0.7064056939501779,
+    "val_f1": 0.759772565742715,
+    "val_macro_f1": 0.6499543100822418
+  }
+]

results/MCFD/stgcn/metrics_test_clean.json

@@ -0,0 +1,14 @@
+{
+  "dataset": "MCFD",
+  "method": "stgcn",
+  "split": "test",
+  "robustness": "clean",
+  "fps": 4238.144418851585,
+  "params": 103394,
+  "accuracy": 0.6793154761904762,
+  "precision": 0.6868456986800182,
+  "recall": 0.8966131907308378,
+  "specificity": 0.3154228855721393,
+  "f1": 0.7778350515463918,
+  "macro_f1": 0.600815921495121
+}

results/MCFD/tcnte/best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5f332d0fd0113260256f02196f93839e283dc00b6042c59fa405d204a4205d0f
+size 851281

results/MCFD/tcnte/history.json

@@ -0,0 +1,519 @@
+[
+  {
+    "epoch": 1,
+    "loss": 0.10085570655454923,
+    "val_fps": 3850.5905707038996,
+    "val_accuracy": 0.511204481792717,
+    "val_precision": 0.9131782945736434,
+    "val_recall": 0.37278481012658227,
+    "val_specificity": 0.900355871886121,
+    "val_f1": 0.529438202247191,
+    "val_macro_f1": 0.5104694653780879
+  },
+  {
+    "epoch": 2,
+    "loss": 0.0920358574052425,
+    "val_fps": 4097.261154995775,
+    "val_accuracy": 0.43790849673202614,
+    "val_precision": 0.9921465968586387,
+    "val_recall": 0.239873417721519,
+    "val_specificity": 0.994661921708185,
+    "val_f1": 0.38634046890927626,
+    "val_macro_f1": 0.43391097519537886
+  },
+  {
+    "epoch": 3,
+    "loss": 0.08832261883009347,
+    "val_fps": 4140.825891191272,
+    "val_accuracy": 0.4911297852474323,
+    "val_precision": 0.9748062015503876,
+    "val_recall": 0.3183544303797468,
+    "val_specificity": 0.9768683274021353,
+    "val_f1": 0.4799618320610687,
+    "val_macro_f1": 0.49089499281298404
+  },
+  {
+    "epoch": 4,
+    "loss": 0.08536651412745175,
+    "val_fps": 3992.569396520161,
+    "val_accuracy": 0.39775910364145656,
+    "val_precision": 0.9898648648648649,
+    "val_recall": 0.18544303797468353,
+    "val_specificity": 0.994661921708185,
+    "val_f1": 0.31236673773987206,
+    "val_macro_f1": 0.3883262260127932
+  },
+  {
+    "epoch": 5,
+    "loss": 0.08266990156193263,
+    "val_fps": 4080.0118669786925,
+    "val_accuracy": 0.4126984126984127,
+    "val_precision": 0.9938650306748467,
+    "val_recall": 0.20506329113924052,
+    "val_specificity": 0.99644128113879,
+    "val_f1": 0.33997901364113325,
+    "val_macro_f1": 0.4054815169130813
+  },
+  {
+    "epoch": 6,
+    "loss": 0.08162742966623943,
+    "val_fps": 4020.9438473275504,
+    "val_accuracy": 0.503734827264239,
+    "val_precision": 0.9905123339658444,
+    "val_recall": 0.330379746835443,
+    "val_specificity": 0.9911032028469751,
+    "val_f1": 0.49549121974371146,
+    "val_macro_f1": 0.5036022933812724
+  },
+  {
+    "epoch": 7,
+    "loss": 0.08149332543582617,
+    "val_fps": 4238.426590306563,
+    "val_accuracy": 0.4715219421101774,
+    "val_precision": 0.9912280701754386,
+    "val_recall": 0.28607594936708863,
+    "val_specificity": 0.9928825622775801,
+    "val_f1": 0.444007858546169,
+    "val_macro_f1": 0.4702245698424795
+  },
+  {
+    "epoch": 8,
+    "loss": 0.07727231089286134,
+    "val_fps": 4093.4655048230225,
+    "val_accuracy": 0.5252100840336135,
+    "val_precision": 0.9577235772357724,
+    "val_recall": 0.37278481012658227,
+    "val_specificity": 0.9537366548042705,
+    "val_f1": 0.5366742596810934,
+    "val_macro_f1": 0.5249192265375309
+  },
+  {
+    "epoch": 9,
+    "loss": 0.07708407426186532,
+    "val_fps": 4077.4532346459714,
+    "val_accuracy": 0.5088702147525677,
+    "val_precision": 0.9258064516129032,
+    "val_recall": 0.3632911392405063,
+    "val_specificity": 0.9181494661921709,
+    "val_f1": 0.5218181818181818,
+    "val_macro_f1": 0.5085098586634096
+  },
+  {
+    "epoch": 10,
+    "loss": 0.07348613941061513,
+    "val_fps": 4178.102348136931,
+    "val_accuracy": 0.5476190476190477,
+    "val_precision": 0.9433962264150944,
+    "val_recall": 0.41139240506329117,
+    "val_specificity": 0.9306049822064056,
+    "val_f1": 0.5729396209784046,
+    "val_macro_f1": 0.5460231603651329
+  },
+  {
+    "epoch": 11,
+    "loss": 0.07192961391747082,
+    "val_fps": 3901.8413411044517,
+    "val_accuracy": 0.5088702147525677,
+    "val_precision": 0.9429530201342282,
+    "val_recall": 0.3556962025316456,
+    "val_specificity": 0.9395017793594306,
+    "val_f1": 0.5165441176470589,
+    "val_macro_f1": 0.5087464421252372
+  },
+  {
+    "epoch": 12,
+    "loss": 0.07018912238428919,
+    "val_fps": 4081.138769960058,
+    "val_accuracy": 0.5196078431372549,
+    "val_precision": 0.946515397082658,
+    "val_recall": 0.369620253164557,
+    "val_specificity": 0.9412811387900356,
+    "val_f1": 0.5316340464269458,
+    "val_macro_f1": 0.519290909174182
+  },
+  {
+    "epoch": 13,
+    "loss": 0.0701279526180488,
+    "val_fps": 3931.66553445532,
+    "val_accuracy": 0.5396825396825397,
+    "val_precision": 0.9486404833836858,
+    "val_recall": 0.39746835443037976,
+    "val_specificity": 0.9395017793594306,
+    "val_f1": 0.56021409455843,
+    "val_macro_f1": 0.5386770766621729
+  },
+  {
+    "epoch": 14,
+    "loss": 0.07136695828825533,
+    "val_fps": 3663.14724819471,
+    "val_accuracy": 0.5513538748832867,
+    "val_precision": 0.8993548387096775,
+    "val_recall": 0.44113924050632913,
+    "val_specificity": 0.8612099644128114,
+    "val_f1": 0.591932059447983,
+    "val_macro_f1": 0.5468732355301087
+  },
+  {
+    "epoch": 15,
+    "loss": 0.06652718700319742,
+    "val_fps": 3711.086271423145,
+    "val_accuracy": 0.5168067226890757,
+    "val_precision": 0.9346092503987241,
+    "val_recall": 0.37088607594936707,
+    "val_specificity": 0.9270462633451957,
+    "val_f1": 0.5310376076121431,
+    "val_macro_f1": 0.5163613651926868
+  },
+  {
+    "epoch": 16,
+    "loss": 0.06348201281094815,
+    "val_fps": 4060.9925181705253,
+    "val_accuracy": 0.5662931839402428,
+    "val_precision": 0.9003690036900369,
+    "val_recall": 0.46329113924050636,
+    "val_specificity": 0.8558718861209964,
+    "val_f1": 0.6117843710823234,
+    "val_macro_f1": 0.5602549565617857
+  },
+  {
+    "epoch": 17,
+    "loss": 0.0651745786265963,
+    "val_fps": 4035.349183937207,
+    "val_accuracy": 0.5709617180205415,
+    "val_precision": 0.9199491740787802,
+    "val_recall": 0.4582278481012658,
+    "val_specificity": 0.8879003558718861,
+    "val_f1": 0.6117448246725813,
+    "val_macro_f1": 0.5661749684134947
+  },
+  {
+    "epoch": 18,
+    "loss": 0.060917166974422086,
+    "val_fps": 4023.0947352137973,
+    "val_accuracy": 0.5723622782446312,
+    "val_precision": 0.8933649289099526,
+    "val_recall": 0.4772151898734177,
+    "val_specificity": 0.8398576512455516,
+    "val_f1": 0.6221122112211221,
+    "val_macro_f1": 0.5648195464707761
+  },
+  {
+    "epoch": 19,
+    "loss": 0.06612300161224459,
+    "val_fps": 4007.706403943383,
+    "val_accuracy": 0.5415499533146592,
+    "val_precision": 0.9164345403899722,
+    "val_recall": 0.41645569620253164,
+    "val_specificity": 0.8932384341637011,
+    "val_f1": 0.5726718885987816,
+    "val_macro_f1": 0.5391053299992901
+  },
+  {
+    "epoch": 20,
+    "loss": 0.06228086257356082,
+    "val_fps": 3946.4649070229502,
+    "val_accuracy": 0.545751633986928,
+    "val_precision": 0.9129251700680272,
+    "val_recall": 0.42468354430379746,
+    "val_specificity": 0.8861209964412812,
+    "val_f1": 0.5796976241900648,
+    "val_macro_f1": 0.5427690761884809
+  },
+  {
+    "epoch": 21,
+    "loss": 0.05918252908262011,
+    "val_fps": 4008.715464844745,
+    "val_accuracy": 0.5606909430438842,
+    "val_precision": 0.895910780669145,
+    "val_recall": 0.45759493670886076,
+    "val_specificity": 0.8505338078291815,
+    "val_f1": 0.6057813154587348,
+    "val_macro_f1": 0.5548674632117079
+  },
+  {
+    "epoch": 22,
+    "loss": 0.05738968637037209,
+    "val_fps": 4041.114440586426,
+    "val_accuracy": 0.5564892623716153,
+    "val_precision": 0.8850855745721271,
+    "val_recall": 0.4582278481012658,
+    "val_specificity": 0.8327402135231317,
+    "val_f1": 0.603836530442035,
+    "val_macro_f1": 0.550062485793658
+  },
+  {
+    "epoch": 23,
+    "loss": 0.0618406231282611,
+    "val_fps": 4056.466660711892,
+    "val_accuracy": 0.5770308123249299,
+    "val_precision": 0.9342783505154639,
+    "val_recall": 0.4588607594936709,
+    "val_specificity": 0.9092526690391459,
+    "val_f1": 0.6154499151103565,
+    "val_macro_f1": 0.5727664513311119
+  },
+  {
+    "epoch": 24,
+    "loss": 0.05784103912188434,
+    "val_fps": 3674.54295708663,
+    "val_accuracy": 0.6055088702147525,
+    "val_precision": 0.8964401294498382,
+    "val_recall": 0.5259493670886076,
+    "val_specificity": 0.8291814946619217,
+    "val_f1": 0.6629437574790586,
+    "val_macro_f1": 0.5937116086213526
+  },
+  {
+    "epoch": 25,
+    "loss": 0.05663871155060647,
+    "val_fps": 3814.9659188790138,
+    "val_accuracy": 0.5798319327731093,
+    "val_precision": 0.912621359223301,
+    "val_recall": 0.4759493670886076,
+    "val_specificity": 0.8718861209964412,
+    "val_f1": 0.6256239600665557,
+    "val_macro_f1": 0.5734502779056183
+  },
+  {
+    "epoch": 26,
+    "loss": 0.055105293738242564,
+    "val_fps": 4110.675356711506,
+    "val_accuracy": 0.5985060690943044,
+    "val_precision": 0.8973509933774835,
+    "val_recall": 0.5145569620253164,
+    "val_specificity": 0.8345195729537367,
+    "val_f1": 0.6540627514078842,
+    "val_macro_f1": 0.5878767594636751
+  },
+  {
+    "epoch": 27,
+    "loss": 0.053442715592788646,
+    "val_fps": 3772.547274277149,
+    "val_accuracy": 0.5513538748832867,
+    "val_precision": 0.9165545087483177,
+    "val_recall": 0.4310126582278481,
+    "val_specificity": 0.8896797153024911,
+    "val_f1": 0.5863108049935428,
+    "val_macro_f1": 0.548127355581932
+  },
+  {
+    "epoch": 28,
+    "loss": 0.05449098952980805,
+    "val_fps": 4097.708849013137,
+    "val_accuracy": 0.5429505135387488,
+    "val_precision": 0.9012016021361816,
+    "val_recall": 0.4272151898734177,
+    "val_specificity": 0.8683274021352313,
+    "val_f1": 0.579647917561185,
+    "val_macro_f1": 0.5394403270670376
+  },
+  {
+    "epoch": 29,
+    "loss": 0.051690250134278894,
+    "val_fps": 4074.740261818418,
+    "val_accuracy": 0.611577964519141,
+    "val_precision": 0.9030172413793104,
+    "val_recall": 0.5303797468354431,
+    "val_specificity": 0.8398576512455516,
+    "val_f1": 0.6682615629984051,
+    "val_macro_f1": 0.5998965472649683
+  },
+  {
+    "epoch": 30,
+    "loss": 0.05760920379900728,
+    "val_fps": 4095.004276021274,
+    "val_accuracy": 0.6274509803921569,
+    "val_precision": 0.8902195608782435,
+    "val_recall": 0.5645569620253165,
+    "val_specificity": 0.8042704626334519,
+    "val_f1": 0.6909372579395817,
+    "val_macro_f1": 0.6110385467136217
+  },
+  {
+    "epoch": 31,
+    "loss": 0.05437249663723008,
+    "val_fps": 3986.0442445740355,
+    "val_accuracy": 0.6171802054154996,
+    "val_precision": 0.8966597077244259,
+    "val_recall": 0.5436708860759494,
+    "val_specificity": 0.8238434163701067,
+    "val_f1": 0.6769109535066982,
+    "val_macro_f1": 0.6036330254360525
+  },
+  {
+    "epoch": 32,
+    "loss": 0.05123214777984182,
+    "val_fps": 4184.958080534249,
+    "val_accuracy": 0.6045751633986928,
+    "val_precision": 0.8903088391906283,
+    "val_recall": 0.529113924050633,
+    "val_specificity": 0.8167259786476868,
+    "val_f1": 0.6637554585152838,
+    "val_macro_f1": 0.5919343864814379
+  },
+  {
+    "epoch": 33,
+    "loss": 0.05163691386569839,
+    "val_fps": 3962.844577665711,
+    "val_accuracy": 0.6022408963585434,
+    "val_precision": 0.9155251141552512,
+    "val_recall": 0.5075949367088608,
+    "val_specificity": 0.8683274021352313,
+    "val_f1": 0.6530944625407166,
+    "val_macro_f1": 0.593505655778017
+  },
+  {
+    "epoch": 34,
+    "loss": 0.051648908634185156,
+    "val_fps": 4176.424816087942,
+    "val_accuracy": 0.6059757236227824,
+    "val_precision": 0.9,
+    "val_recall": 0.5240506329113924,
+    "val_specificity": 0.8362989323843416,
+    "val_f1": 0.6624,
+    "val_macro_f1": 0.5946529147982063
+  },
+  {
+    "epoch": 35,
+    "loss": 0.04855069010976845,
+    "val_fps": 3658.422542279308,
+    "val_accuracy": 0.5732959850606909,
+    "val_precision": 0.8854166666666666,
+    "val_recall": 0.48417721518987344,
+    "val_specificity": 0.8238434163701067,
+    "val_f1": 0.6260229132569558,
+    "val_macro_f1": 0.5646418914110867
+  },
+  {
+    "epoch": 36,
+    "loss": 0.04994714245287842,
+    "val_fps": 3890.497311736381,
+    "val_accuracy": 0.5933706816059757,
+    "val_precision": 0.9079401611047181,
+    "val_recall": 0.49936708860759493,
+    "val_specificity": 0.8576512455516014,
+    "val_f1": 0.6443446304614128,
+    "val_macro_f1": 0.5848426149582269
+  },
+  {
+    "epoch": 37,
+    "loss": 0.04879177145303869,
+    "val_fps": 3921.6370564224535,
+    "val_accuracy": 0.6335200746965453,
+    "val_precision": 0.8388746803069054,
+    "val_recall": 0.6227848101265823,
+    "val_specificity": 0.6637010676156584,
+    "val_f1": 0.7148565201598256,
+    "val_macro_f1": 0.6010598734045373
+  },
+  {
+    "epoch": 38,
+    "loss": 0.05096233196253656,
+    "val_fps": 4071.6557654294024,
+    "val_accuracy": 0.6251167133520075,
+    "val_precision": 0.8904522613065327,
+    "val_recall": 0.5607594936708861,
+    "val_specificity": 0.806049822064057,
+    "val_f1": 0.6881553398058252,
+    "val_macro_f1": 0.6091449607162538
+  },
+  {
+    "epoch": 39,
+    "loss": 0.0486308291618222,
+    "val_fps": 4121.749462498118,
+    "val_accuracy": 0.6078431372549019,
+    "val_precision": 0.8911205073995772,
+    "val_recall": 0.5335443037974683,
+    "val_specificity": 0.8167259786476868,
+    "val_f1": 0.667458432304038,
+    "val_macro_f1": 0.5948213663226675
+  },
+  {
+    "epoch": 40,
+    "loss": 0.047241462193982155,
+    "val_fps": 3984.0521283028124,
+    "val_accuracy": 0.626984126984127,
+    "val_precision": 0.898876404494382,
+    "val_recall": 0.5569620253164557,
+    "val_specificity": 0.8238434163701067,
+    "val_f1": 0.687768659632669,
+    "val_macro_f1": 0.6122901269177838
+  },
+  {
+    "epoch": 41,
+    "loss": 0.052161689864979065,
+    "val_fps": 4086.749805035564,
+    "val_accuracy": 0.5648926237161531,
+    "val_precision": 0.9263157894736842,
+    "val_recall": 0.44556962025316454,
+    "val_specificity": 0.900355871886121,
+    "val_f1": 0.6017094017094017,
+    "val_macro_f1": 0.5611427666983222
+  },
+  {
+    "epoch": 42,
+    "loss": 0.05344774344067133,
+    "val_fps": 4152.020338004575,
+    "val_accuracy": 0.6213818860877685,
+    "val_precision": 0.9017763845350052,
+    "val_recall": 0.5462025316455696,
+    "val_specificity": 0.8327402135231317,
+    "val_f1": 0.6803310997240836,
+    "val_macro_f1": 0.6080533575323337
+  },
+  {
+    "epoch": 43,
+    "loss": 0.05571009869821444,
+    "val_fps": 3913.0392604681992,
+    "val_accuracy": 0.607376283846872,
+    "val_precision": 0.8893572181243414,
+    "val_recall": 0.5341772151898734,
+    "val_specificity": 0.8131672597864769,
+    "val_f1": 0.6674574930802689,
+    "val_macro_f1": 0.5941276069389949
+  },
+  {
+    "epoch": 44,
+    "loss": 0.05139429723915675,
+    "val_fps": 4071.552554794066,
+    "val_accuracy": 0.6213818860877685,
+    "val_precision": 0.9017763845350052,
+    "val_recall": 0.5462025316455696,
+    "val_specificity": 0.8327402135231317,
+    "val_f1": 0.6803310997240836,
+    "val_macro_f1": 0.6080533575323337
+  },
+  {
+    "epoch": 45,
+    "loss": 0.049104086001179104,
+    "val_fps": 4092.964916941528,
+    "val_accuracy": 0.6358543417366946,
+    "val_precision": 0.8780718336483931,
+    "val_recall": 0.5879746835443038,
+    "val_specificity": 0.7704626334519573,
+    "val_f1": 0.7043214556482184,
+    "val_macro_f1": 0.6152226962323717
+  },
+  {
+    "epoch": 46,
+    "loss": 0.047514633581137944,
+    "val_fps": 4214.2100619906305,
+    "val_accuracy": 0.6321195144724556,
+    "val_precision": 0.9057377049180327,
+    "val_recall": 0.5594936708860759,
+    "val_specificity": 0.8362989323843416,
+    "val_f1": 0.6917057902973396,
+    "val_macro_f1": 0.6178436358894106
+  },
+  {
+    "epoch": 47,
+    "loss": 0.04549025025912235,
+    "val_fps": 4043.120703372622,
+    "val_accuracy": 0.6092436974789915,
+    "val_precision": 0.8841778697001034,
+    "val_recall": 0.5411392405063291,
+    "val_specificity": 0.800711743772242,
+    "val_f1": 0.6713780918727915,
+    "val_macro_f1": 0.5947564034493491
+  }
+]

results/MCFD/tcnte/metrics_test_clean.json

@@ -0,0 +1,14 @@
+{
+  "dataset": "MCFD",
+  "method": "tcnte",
+  "split": "test",
+  "robustness": "clean",
+  "fps": 3795.2966462786967,
+  "params": 210242,
+  "accuracy": 0.5829613095238095,
+  "precision": 0.7027417027417028,
+  "recall": 0.5787284610814023,
+  "specificity": 0.5900497512437811,
+  "f1": 0.634734441186054,
+  "macro_f1": 0.5744110003936338
+}

results/Synthetic/agcn/best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:de34d04634cd958d5a365026da64387139ba25e4183b2b1334f26a61c5ac0c74
+size 867809

results/Synthetic/agcn/history.json

@@ -0,0 +1,13 @@
+[
+  {
+    "epoch": 1,
+    "loss": 0.1952398121356964,
+    "val_fps": 145.72827408063176,
+    "val_accuracy": 0.4444444444444444,
+    "val_precision": 0.4444444444444444,
+    "val_recall": 1.0,
+    "val_specificity": 0.0,
+    "val_f1": 0.6153846153846154,
+    "val_macro_f1": 0.3076923076923077
+  }
+]

results/Synthetic/agcn/metrics_test_clean.json

@@ -0,0 +1,14 @@
+{
+  "dataset": "Synthetic",
+  "method": "agcn",
+  "split": "test",
+  "robustness": "clean",
+  "fps": 159.77018381668333,
+  "params": 208520,
+  "accuracy": 0.45454545454545453,
+  "precision": 0.45454545454545453,
+  "recall": 1.0,
+  "specificity": 0.0,
+  "f1": 0.625,
+  "macro_f1": 0.3125
+}

results/Synthetic/ctrgcn/best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:881a9adb7a44a4a9f5130a0e0f5155cd9218cb0a3d981e08b7bbd750a780ed2f
+size 1535165

results/Synthetic/ctrgcn/history.json

@@ -0,0 +1,13 @@
+[
+  {
+    "epoch": 1,
+    "loss": 0.19207785030206045,
+    "val_fps": 135.12070967593993,
+    "val_accuracy": 0.4444444444444444,
+    "val_precision": 0.4444444444444444,
+    "val_recall": 1.0,
+    "val_specificity": 0.0,
+    "val_f1": 0.6153846153846154,
+    "val_macro_f1": 0.3076923076923077
+  }
+]

results/Synthetic/ctrgcn/metrics_test_clean.json

@@ -0,0 +1,14 @@
+{
+  "dataset": "Synthetic",
+  "method": "ctrgcn",
+  "split": "test",
+  "robustness": "clean",
+  "fps": 151.245780615558,
+  "params": 374984,
+  "accuracy": 0.45454545454545453,
+  "precision": 0.45454545454545453,
+  "recall": 1.0,
+  "specificity": 0.0,
+  "f1": 0.625,
+  "macro_f1": 0.3125
+}

results/Synthetic/dynafall/best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:21fe350dd65783d64f108c12a90cba472ea1652d3c02a287cac988dba799441f
+size 1265049

results/Synthetic/dynafall/history.json

@@ -0,0 +1,13 @@
+[
+  {
+    "epoch": 1,
+    "loss": 0.22574210166931152,
+    "val_fps": 127.35835051918123,
+    "val_accuracy": 0.4444444444444444,
+    "val_precision": 0.4444444444444444,
+    "val_recall": 1.0,
+    "val_specificity": 0.0,
+    "val_f1": 0.6153846153846154,
+    "val_macro_f1": 0.3076923076923077
+  }
+]

results/Synthetic/dynafall/metrics_robust_clean.json

@@ -0,0 +1,14 @@
+{
+  "dataset": "Synthetic",
+  "method": "dynafall",
+  "split": "test",
+  "robustness": "clean",
+  "fps": 112.37041093419398,
+  "params": 307017,
+  "accuracy": 0.45454545454545453,
+  "precision": 0.45454545454545453,
+  "recall": 1.0,
+  "specificity": 0.0,
+  "f1": 0.625,
+  "macro_f1": 0.3125
+}

results/Synthetic/dynafall/metrics_robust_low_conf.json

@@ -0,0 +1,14 @@
+{
+  "dataset": "Synthetic",
+  "method": "dynafall",
+  "split": "test",
+  "robustness": "low_conf",
+  "fps": 199.52054000732969,
+  "params": 307017,
+  "accuracy": 0.45454545454545453,
+  "precision": 0.45454545454545453,
+  "recall": 1.0,
+  "specificity": 0.0,
+  "f1": 0.625,
+  "macro_f1": 0.3125
+}

results/Synthetic/dynafall/metrics_robust_lower_body.json

@@ -0,0 +1,14 @@
+{
+  "dataset": "Synthetic",
+  "method": "dynafall",
+  "split": "test",
+  "robustness": "lower_body",
+  "fps": 199.43313977667506,
+  "params": 307017,
+  "accuracy": 0.45454545454545453,
+  "precision": 0.45454545454545453,
+  "recall": 1.0,
+  "specificity": 0.0,
+  "f1": 0.625,
+  "macro_f1": 0.3125
+}

results/Synthetic/dynafall/metrics_robust_missing_10.json

@@ -0,0 +1,14 @@
+{
+  "dataset": "Synthetic",
+  "method": "dynafall",
+  "split": "test",
+  "robustness": "missing",
+  "fps": 204.83560277908077,
+  "params": 307017,
+  "accuracy": 0.45454545454545453,
+  "precision": 0.45454545454545453,
+  "recall": 1.0,
+  "specificity": 0.0,
+  "f1": 0.625,
+  "macro_f1": 0.3125
+}

results/Synthetic/dynafall/metrics_robust_missing_20.json

@@ -0,0 +1,14 @@
+{
+  "dataset": "Synthetic",
+  "method": "dynafall",
+  "split": "test",
+  "robustness": "missing",
+  "fps": 201.4036998642287,
+  "params": 307017,
+  "accuracy": 0.45454545454545453,
+  "precision": 0.45454545454545453,
+  "recall": 1.0,
+  "specificity": 0.0,
+  "f1": 0.625,
+  "macro_f1": 0.3125
+}

results/Synthetic/dynafall/metrics_robust_missing_30.json

@@ -0,0 +1,14 @@
+{
+  "dataset": "Synthetic",
+  "method": "dynafall",
+  "split": "test",
+  "robustness": "missing",
+  "fps": 202.57249108682828,
+  "params": 307017,
+  "accuracy": 0.45454545454545453,
+  "precision": 0.45454545454545453,
+  "recall": 1.0,
+  "specificity": 0.0,
+  "f1": 0.625,
+  "macro_f1": 0.3125
+}

results/Synthetic/dynafall/metrics_robust_upper_body.json

@@ -0,0 +1,14 @@
+{
+  "dataset": "Synthetic",
+  "method": "dynafall",
+  "split": "test",
+  "robustness": "upper_body",
+  "fps": 201.61263794877428,
+  "params": 307017,
+  "accuracy": 0.45454545454545453,
+  "precision": 0.45454545454545453,
+  "recall": 1.0,
+  "specificity": 0.0,
+  "f1": 0.625,
+  "macro_f1": 0.3125
+}

results/Synthetic/dynafall/metrics_test_clean.json

@@ -0,0 +1,14 @@
+{
+  "dataset": "Synthetic",
+  "method": "dynafall",
+  "split": "test",
+  "robustness": "clean",
+  "fps": 137.54780410253272,
+  "params": 307017,
+  "accuracy": 0.45454545454545453,
+  "precision": 0.45454545454545453,
+  "recall": 1.0,
+  "specificity": 0.0,
+  "f1": 0.625,
+  "macro_f1": 0.3125
+}

results/Synthetic/lstm/best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:14fb82f3d7fa097b915714d19994e29d2f7741534949e898b1f637d0f70b18e9
+size 462662