Fine-tuned Faster R-CNN ResNet50-FPN on loc_beyond_words

README.md

@@ -0,0 +1,75 @@
+---
+license: bsd-3-clause
+language:
+  - en
+tags:
+  - object-detection
+  - faster-rcnn
+  - resnet50
+  - document-layout
+  - historical-newspapers
+pipeline_tag: object-detection
+metrics:
+  - 0.3251
+widget:
+  - src: https://datasets-server.huggingface.co/cached-assets/biglam/loc_beyond_words/--/6c7f5fb3c60f02d9fe925cfc14aa7008f6c89099/--/default/train/0/image/image.jpg
+---
+
+# opencode-r1
+
+Object detection model fine-tuned from **torchvision Faster R-CNN (ResNet-50-FPN)**
+pre-trained on COCO (base model license: BSD-3-Clause, open and shareable) on the
+[`biglam/loc_beyond_words`](https://huggingface.co/datasets/biglam/loc_beyond_words) dataset
+(Library of Congress "Beyond Words", data license CC0-1.0).
+
+## Classes (7) + background
+
+Photograph, Illustration, Map, Comics/Cartoon, Editorial Cartoon, Headline, Advertisement
+
+## Validation results (COCO-style, biglam/loc_beyond_words validation set)
+
+| Metric | Value |
+|---|---|
+| mAP @[IoU=0.50:0.95] | 0.3251 |
+| mAP @ IoU=0.50 | 0.4175 |
+| mAP @ IoU=0.75 | 0.3737 |
+
+## Training
+
+| Setting | Value |
+|---|---|
+| Base model | Faster R-CNN ResNet50-FPN (COCO, BSD-3-Clause) |
+| Epochs | 12 |
+| Batch size | 8 |
+| Max image dim | 520 |
+| Optimizer | SGD (momentum 0.9), StepLR x0.3/8 epochs |
+| Head LR / Backbone LR | 0.002 / 0.0002 |
+| Hardware | NVIDIA GPU (Hugging Face jobs) |
+
+Images are downscaled so the largest dimension is 520px (aspect preserved);
+boxes scaled accordingly. Predictions below score 0.5 are discarded.
+
+## To load and run
+
+```python
+import torch
+from torchvision.models.detection import fasterrcnn_resnet50_fpn
+from torchvision.models.detection.faster_rcnn import FastRCNNPredictor
+from huggingface_hub import hf_hub_download
+from PIL import Image
+import numpy as np
+
+state = torch.load(hf_hub_download("harness-race/opencode-r1", "model.pth"),
+                   map_location="cpu")
+model = fasterrcnn_resnet50_fpn(weights=None)
+in_features = model.roi_heads.box_predictor.cls_score.in_features
+model.roi_heads.box_predictor = FastRCNNPredictor(in_features, 1 + 7)  # 7 + background
+model.load_state_dict(state)
+model.eval()
+
+img = Image.open("page.jpg").convert("RGB")
+# resize to max-dim 520 like training, then:
+x = torch.as_tensor(np.asarray(img), dtype=torch.float32).permute(2, 0, 1) / 255.0
+with torch.no_grad():
+    dets = model([x])[0]
+```

args.json

@@ -0,0 +1 @@
+{"epochs": 12, "batch": 8, "max_dim": 520, "lr": 0.002, "base_lr": 0.0002, "seed": 0, "push": "1"}

class_names.json

@@ -0,0 +1 @@
+["Photograph", "Illustration", "Map", "Comics/Cartoon", "Editorial Cartoon", "Headline", "Advertisement"]

full_model.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:39a92f7e53f5685d1b5da785d055dc26340d3aadcd793b0da8bbafa378f152f4
+size 165882301

model.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:77a36019e9a77eb115c22db30856bf20137cd7b93641c58247a246cc8bdd66a8
+size 165831383

results.json

@@ -0,0 +1 @@
+{"mAP_050_095": 0.325054090182978, "mAP_050": 0.41747069516008106, "mAP_075": 0.373689980761858}