WALDO (code, models)

.gitattributes

@@ -33,3 +33,9 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+models/waldo30/WALDO30_yolov8m_640x640_confusion_matrix_normalized.png filter=lfs diff=lfs merge=lfs -text
+models/waldo30/WALDO30_yolov8m_640x640_F1_curve.png filter=lfs diff=lfs merge=lfs -text
+models/waldo30/WALDO30_yolov8n_416x416_confusion_matrix_normalized.png filter=lfs diff=lfs merge=lfs -text
+models/waldo30/WALDO30_yolov8n_416x416_F1_curve.pngF1_curve.png filter=lfs diff=lfs merge=lfs -text
+models/waldo30/WALDO30_yolov8n_640x640_confusion_matrix_normalized_F1_curve.png filter=lfs diff=lfs merge=lfs -text
+models/waldo30/WALDO30_yolov8n_640x640_confusion_matrix_normalized.png filter=lfs diff=lfs merge=lfs -text

code/WALDO.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4e67e13a15d9d7c735afdf5fe69daffc3b7159390aa81dddfe1403551fc4704d
+size 169975955

models/waldo30-onnx/.gitattributes

@@ -0,0 +1,35 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text

models/waldo30-onnx/config.json

@@ -0,0 +1,31 @@
+{
+    "model_type": "yolos",
+    "id2label": {
+      "0": "LightVehicle",
+      "1": "Person",
+      "2": "Building",
+      "3": "UPole",
+      "4": "Boat",
+      "5": "Bike",
+      "6": "Container",
+      "7": "Truck",
+      "8": "Gastank",
+      "9": "Digger",
+      "10": "SolarPanels",
+      "11": "Bus"
+    },
+    "label2id": {
+      "LightVehicle": 0,
+      "Person": 1,
+      "Building": 2,
+      "UPole": 3,
+      "Boat": 4,
+      "Bike": 5,
+      "Container": 6,
+      "Truck": 7,
+      "Gastank": 8,
+      "Digger": 9,
+      "SolarPanels": 10,
+      "Bus": 11
+    }
+  }

models/waldo30-onnx/onnx/model.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:be80223d0d59f36baffb9542b52b087dee3712c8e8d120b65453576b69796c27
+size 103632215

models/waldo30-onnx/preprocessor_config.json

@@ -0,0 +1,13 @@
+{
+    "do_normalize": false,
+    "do_pad": false,
+    "do_rescale": true,
+    "do_resize": true,
+    "feature_extractor_type": "ImageFeatureExtractor",
+    "resample": 2,
+    "rescale_factor": 0.00392156862745098,
+    "size": {
+      "width": 640,
+      "height": 640
+    }
+}

models/waldo30-onnx/source.txt

@@ -0,0 +1 @@
+https://huggingface.co/mhassanch/waldo30-onnx

models/waldo30-yolov8m-640x640/.gitattributes

@@ -0,0 +1,35 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text

models/waldo30-yolov8m-640x640/README.md

@@ -0,0 +1,10 @@
+---
+base_model:
+- StephanST/WALDO30
+tags:
+- onnx
+- yolov8
+- transformer.js
+- '@geobase-js/geoai'
+---
+This ONNX model is a converted version of the original model available at: [StephanST/WALDO30](https://huggingface.co/StephanST/WALDO30).

models/waldo30-yolov8m-640x640/config.json

@@ -0,0 +1,31 @@
+{
+    "model_type": "yolos",
+    "id2label": {
+      "0": "LightVehicle",
+      "1": "Person",
+      "2": "Building",
+      "3": "UPole",
+      "4": "Boat",
+      "5": "Bike",
+      "6": "Container",
+      "7": "Truck",
+      "8": "Gastank",
+      "9": "Digger",
+      "10": "SolarPanels",
+      "11": "Bus"
+    },
+    "label2id": {
+      "LightVehicle": 0,
+      "Person": 1,
+      "Building": 2,
+      "UPole": 3,
+      "Boat": 4,
+      "Bike": 5,
+      "Container": 6,
+      "Truck": 7,
+      "Gastank": 8,
+      "Digger": 9,
+      "SolarPanels": 10,
+      "Bus": 11
+    }
+  }

models/waldo30-yolov8m-640x640/onnx/model.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:46df6260eb4a7ef47e5f625985bd5f4f0616b13020f26ca55932171c0c44f074
+size 103632215

models/waldo30-yolov8m-640x640/onnx/model_quantized.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:46df6260eb4a7ef47e5f625985bd5f4f0616b13020f26ca55932171c0c44f074
+size 103632215

models/waldo30-yolov8m-640x640/preprocessor_config.json

@@ -0,0 +1,13 @@
+{
+    "do_normalize": false,
+    "do_pad": false,
+    "do_rescale": true,
+    "do_resize": true,
+    "feature_extractor_type": "ImageFeatureExtractor",
+    "resample": 2,
+    "rescale_factor": 0.00392156862745098,
+    "size": {
+      "width": 640,
+      "height": 640
+    }
+}

models/waldo30/.gitattributes

@@ -0,0 +1,35 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text

models/waldo30/README.md

@@ -0,0 +1,175 @@
+---
+language:
+- en
+base_model:
+- Ultralytics/YOLOv8
+pipeline_tag: object-detection
+---
+
+W.A.L.D.O.
+Whereabouts Ascertainment for Low-lying Detectable Objects 
+---------------------------------------------------------------------
+
+
+[![WALDO 3.0 preview vid](https://i.imgur.com/hGghrLn.jpeg)](https://www.youtube.com/watch?v=1y5y9yklj2U)
+
+Welcome to the WALDO v3.0 public release
+---------------------------------------------------------------------
+
+
+WHAT IS WALDO?
+
+WALDO is a detection AI model, based on a large YOLO-v8 backbone and my own
+synthetic data pipeline. **The model is capable of detecting these classes 
+of items in overhead imagery ranging in altitude from about 30 feet to 
+satellite imagery!**
+
+
+Output classes:
+
+0 -> 'LightVehicle'  --> all kinds of civilan cars, including pickup trucks, vans etc... 🚗🏎️🚓🚐🚑 </br>
+1 -> 'Person' --> people! all kinds of people including ones that are on bikes or swimming in the sea 🧍‍♀️🕺💃🧜🏽‍♀️🏂🧞</br> 
+2 -> 'Building' --> all kinds of buildings 🕌🏛️🏭🏡</br>
+3 -> 'UPole' --> utility poles, power poles, anything thin and sticking up that you should avoid with a drone 🎏</br>
+4 -> 'Boat' --> boats, ships, canoes, kayaks, surf boards... all the floaty stuff 🚢🏄</br>
+5 -> 'Bike' --> bikes, mopeds, motorbikes, all stuff with 2 wheels 🚲</br>
+6 -> 'Container' --> shipping containers, including on the back of an articulated truck... 📦🏗️</br>
+7 -> 'Truck' --> large commercial vehicles including articulated trucks or big box-on-chassis delivery trucks 🚚</br>
+8 -> 'Gastank'--> cylindrical tanks such as butane tanks and gas expansion tanks, or grain silos... pretty much anything that looks cylindrical for storing liquids 🫙</br>
+10 -> 'Digger' --> all kinds of construction vehicles, including tractors and construction gear 🚜</br>
+11 -> 'Solarpanels' --> solar panels ▪️🌞▪️</br>
+12 -> 'Bus' --> a bus 🚌</br>
+
+--> In general the lower the class number the better-trained you can expect it to be.
+For users of previous versions of WALDO: note that I removed the military class and smoke detection. This is meant to be a FOSS tool for civilian use and I don't want to pursue making it work for military applications.
+
+
+---------------------------------------------------------------------
+
+WHERE IS WALDO?
+
+Right here on HF!
+
+Note there are a couple more models that have slightly better performance over on Gumroad here: https://6228189440665.gumroad.com/l/WALDOv3
+Those are for sale as a kind of sponsorship for the project: if you find value in the free ones here you can buy those for a nice little performance boost... but it's entirey up to you! 
+
+
+[![P2 model performance boost](https://i.imgur.com/VKa5NN5.png)]
+
+
+In both cases the actual files are MIT license and you can freely share them, so if someone gives you the ones from Gumroad you are free yo use them including commercially. It's really just a way to offset some of the work and compute that went into making this project and keeping it FOSS.
+
+
+---------------------------------------------------------------------                                                                                                                                                               
+
+WHAT IS IT GOOD FOR?
+
+People are currently using versions of WALDO for:
+1. disaster recovery
+2. monitoring wildlife sanctuaries (intruder detection)
+3. occupancy calculation (parking lots etc..)
+4. monitoring infrastructure 
+5. construction site monitoring
+6. traffic flow management
+7. crowd counting
+8. some fun AI art applications!
+9. drone safety (avoiding people / cars on the ground)
+10. lots of other fun stuff...
+
+The main reason for me to make WALDO free has in fact been discovering all these cool applications. Let me know what you build!
+
+---------------------------------------------------------------------                                                                                                                                                               
+
+FOR AI NERDS !
+
+It's a set of YOLOv8 model, trained on my own datasets of synthetic and "augmented" / semi-synthetic data.
+I'm not going to release the dataset for the time being.
+
+The weights are completely open, allowing you to deploy in any number of ways this time! 
+
+
+---------------------------------------------------------------------                                                                                
+
+HOW CAN I START WITH WALDO?  
+
+Check out the boilerplate code in the repo to run the models and output pretty detections using the wonderful Supervision annotation library from Roboflow :) 
+
+---------------------------------------------------------------------
+
+GOING DEEPER
+
+Of course if you know your way around deploying AI models there is a lot more you do
+with this release, inclusing:
+
+1. fine-tuning the models on your own data (if you know what you are doing, this is probably your starting point)
+2. building a nicely optimized sliding-window inference setup that works nicely on your edge hardware
+3. quantizing the models for super-duper edge performance on cheap devices
+4. using the models to annotate your own data and train something of your own!
+
+
+Enjoy!
+
+---------------------------------------------------------------------
+
+
+PREVIOUS VERSIONS
+
+I am retiring the old versions, this is the only one that will stay online.
+
+---------------------------------------------------------------------
+
+
+CAN YOU HELP ME WITH X? 
+
+Sure, email me at stephan.sturges@gmail.com
+
+
+---------------------------------------------------------------------
+
+
+DETECTION OF X ISN'T WORKING AS EXPECTED:
+
+I'd love to see example images, videos, sample data, etc at:
+stephan.sturges@gmail.com
+
+
+---------------------------------------------------------------------
+
+SUPPORT WALDO!
+
+Visit [![the WALDO gumroad page](https://t.co/kRvhYkVxW2)] to support the project!
+
+---------------------------------------------------------------------
+
+
+LICENSE
+----------------------------------------------------------------------------
+
+Unless otherwise specified all code in this release is published with the 
+licence conditions below.
+----------------------------------------------------------------------------
+
+
+MIT License
+
+Copyright (c) 2024 Stephan Sturges / Aircortex.com 
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+The software cannot be used for the 2025 SPRIND "fully autonomous flight 2.0" competition.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

models/waldo30/WALDO30_yolov8m_640x640.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0f1ee3daa755f68ffe1eea7a51ab6074614bb38df2097587317548b0395064da
+size 52075265

models/waldo30/WALDO30_yolov8n_416x416.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ae339e0dfff2c2e9652244e341ae7117bbae2d66d1f7c39c15fce64f25802fd1
+size 6254489

models/waldo30/WALDO30_yolov8n_640x640.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:580cd0ee6e120808b5848ca9349717e280a0124ad5365cebee74c8fbd8c1881b
+size 6285145

models/waldo30/run_sliced_inference.py

@@ -0,0 +1,116 @@
+import cv2
+import sys
+from sahi.models.yolov8 import Yolov8DetectionModel
+from sahi.predict import get_sliced_prediction
+import supervision as sv
+import numpy as np
+
+# Check the number of command-line arguments
+if len(sys.argv) != 8:
+    print("Usage: python yolov8_video_inference.py <model_path> <input_video_path> <output_video_path> <slice_height> <slice_width> <overlap_height_ratio> <overlap_width_ratio>")
+    sys.exit(1)
+
+# Get command-line arguments
+model_path = sys.argv[1]
+input_video_path = sys.argv[2]
+output_video_path = sys.argv[3]
+slice_height = int(sys.argv[4])
+slice_width = int(sys.argv[5])
+overlap_height_ratio = float(sys.argv[6])
+overlap_width_ratio = float(sys.argv[7])
+
+# Load YOLOv8 model with SAHI
+detection_model = Yolov8DetectionModel(
+    model_path=model_path,
+    confidence_threshold=0.1,
+    device="cuda"  # or "cpu"
+)
+
+# Open input video
+cap = cv2.VideoCapture(input_video_path)
+width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
+height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
+fps = cap.get(cv2.CAP_PROP_FPS)
+fourcc = cv2.VideoWriter_fourcc(*"mp4v")
+
+# Set up output video writer
+out = cv2.VideoWriter(output_video_path, fourcc, fps, (width, height))
+
+# Create bounding box and label annotators
+#box_annotator = sv.BoundingBoxAnnotator(thickness=1)
+box_annotator = sv.BoxCornerAnnotator(thickness=2)
+label_annotator = sv.LabelAnnotator(text_scale=0.5, text_thickness=2)
+
+# Process each frame
+frame_count = 0
+while cap.isOpened():
+    ret, frame = cap.read()
+    if not ret:
+        break
+
+    # Perform sliced inference on the current frame using SAHI
+    result = get_sliced_prediction(
+        image=frame,
+        detection_model=detection_model,
+        slice_height=slice_height,
+        slice_width=slice_width,
+        overlap_height_ratio=overlap_height_ratio,
+        overlap_width_ratio=overlap_width_ratio
+    )
+
+    # Extract data from SAHI result
+    object_predictions = result.object_prediction_list
+
+    # Initialize lists to hold the data
+    xyxy = []
+    confidences = []
+    class_ids = []
+    class_names = []
+
+    # Loop over the object predictions and extract data
+    for pred in object_predictions:
+        bbox = pred.bbox.to_xyxy()  # Convert bbox to [x1, y1, x2, y2]
+        xyxy.append(bbox)
+        confidences.append(pred.score.value)
+        class_ids.append(pred.category.id)
+        class_names.append(pred.category.name)
+
+    # Check if there are any detections
+    if xyxy:
+        # Convert lists to numpy arrays
+        xyxy = np.array(xyxy, dtype=np.float32)
+        confidences = np.array(confidences, dtype=np.float32)
+        class_ids = np.array(class_ids, dtype=int)
+
+        # Create sv.Detections object
+        detections = sv.Detections(
+            xyxy=xyxy,
+            confidence=confidences,
+            class_id=class_ids
+        )
+
+        # Prepare labels for label annotator
+        labels = [
+            f"{class_name} {confidence:.2f}"
+            for class_name, confidence in zip(class_names, confidences)
+        ]
+
+        # Annotate frame with detection results
+        annotated_frame = frame.copy()
+        annotated_frame = box_annotator.annotate(scene=annotated_frame, detections=detections)
+        annotated_frame = label_annotator.annotate(scene=annotated_frame, detections=detections, labels=labels)
+    else:
+        # If no detections, use the original frame
+        annotated_frame = frame.copy()
+
+    # Write the annotated frame to the output video
+    out.write(annotated_frame)
+
+    frame_count += 1
+    print(f"Processed frame {frame_count}", end='\r')
+
+# Release resources
+cap.release()
+out.release()
+print("\nInference complete. Video saved at", output_video_path)
+

models/waldo30/run_sliced_inference_with_tracker.py

@@ -0,0 +1,152 @@
+import cv2
+import sys
+from sahi.models.yolov8 import Yolov8DetectionModel
+from sahi.predict import get_sliced_prediction
+import supervision as sv
+import numpy as np
+
+# Check the number of command-line arguments
+if len(sys.argv) != 8:
+    print("Usage: python yolov8_video_inference.py <model_path> <input_video_path> <output_video_path> <slice_height> <slice_width> <overlap_height_ratio> <overlap_width_ratio>")
+    sys.exit(1)
+
+# Get command-line arguments
+model_path = sys.argv[1]
+input_video_path = sys.argv[2]
+output_video_path = sys.argv[3]
+slice_height = int(sys.argv[4])
+slice_width = int(sys.argv[5])
+overlap_height_ratio = float(sys.argv[6])
+overlap_width_ratio = float(sys.argv[7])
+
+# Load YOLOv8 model with SAHI
+detection_model = Yolov8DetectionModel(
+    model_path=model_path,
+    confidence_threshold=0.25,
+    device="cuda"  # or "cpu"
+)
+
+# Get video info
+video_info = sv.VideoInfo.from_video_path(video_path=input_video_path)
+
+# Open input video
+cap = cv2.VideoCapture(input_video_path)
+width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
+height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
+fps = cap.get(cv2.CAP_PROP_FPS)
+fourcc = cv2.VideoWriter_fourcc(*"mp4v")
+
+# Set up output video writer
+out = cv2.VideoWriter(output_video_path, fourcc, fps, (width, height))
+
+# Initialize tracker and smoother
+tracker = sv.ByteTrack(frame_rate=video_info.fps)
+smoother = sv.DetectionsSmoother()
+
+# Create bounding box and label annotators
+box_annotator = sv.BoxCornerAnnotator(thickness=2)
+label_annotator = sv.LabelAnnotator(
+    text_scale=0.5,
+    text_thickness=1,
+    text_padding=1
+)
+
+# Process each frame
+frame_count = 0
+class_id_to_name = {}  # Initialize once to store class_id to name mapping
+
+while cap.isOpened():
+    ret, frame = cap.read()
+    if not ret:
+        break
+
+    # Perform sliced inference on the current frame using SAHI
+    result = get_sliced_prediction(
+        image=frame,
+        detection_model=detection_model,
+        slice_height=slice_height,
+        slice_width=slice_width,
+        overlap_height_ratio=overlap_height_ratio,
+        overlap_width_ratio=overlap_width_ratio
+    )
+
+    # Extract data from SAHI result
+    object_predictions = result.object_prediction_list
+
+    # Initialize lists to hold the data
+    xyxy = []
+    confidences = []
+    class_ids = []
+    # Build or update class_id to name mapping
+    for pred in object_predictions:
+        if pred.category.id not in class_id_to_name:
+            class_id_to_name[pred.category.id] = pred.category.name
+
+    # Loop over the object predictions and extract data
+    for pred in object_predictions:
+        bbox = pred.bbox.to_xyxy()  # Convert bbox to [x1, y1, x2, y2]
+        xyxy.append(bbox)
+        confidences.append(pred.score.value)
+        class_ids.append(pred.category.id)
+
+    # Check if there are any detections
+    if xyxy:
+        # Convert lists to numpy arrays
+        xyxy = np.array(xyxy, dtype=np.float32)
+        confidences = np.array(confidences, dtype=np.float32)
+        class_ids = np.array(class_ids, dtype=int)
+
+        # Create sv.Detections object
+        detections = sv.Detections(
+            xyxy=xyxy,
+            confidence=confidences,
+            class_id=class_ids
+        )
+
+        # Update tracker with detections
+        detections = tracker.update_with_detections(detections)
+
+        # Update smoother with detections
+        detections = smoother.update_with_detections(detections)
+
+        # Prepare labels for label annotator
+        # Include tracker ID in labels if available
+        labels = []
+        for i in range(len(detections.xyxy)):
+            class_id = detections.class_id[i]
+            confidence = detections.confidence[i]
+            class_name = class_id_to_name.get(class_id, 'Unknown')
+            label = f"{class_name} {confidence:.2f}"
+
+            # Add tracker ID if available
+            if hasattr(detections, 'tracker_id') and detections.tracker_id is not None:
+                tracker_id = detections.tracker_id[i]
+                label = f"ID {tracker_id} {label}"
+
+            labels.append(label)
+
+        # Annotate frame with detection results
+        annotated_frame = frame.copy()
+        annotated_frame = box_annotator.annotate(
+            scene=annotated_frame,
+            detections=detections
+        )
+        annotated_frame = label_annotator.annotate(
+            scene=annotated_frame,
+            detections=detections,
+            labels=labels
+        )
+    else:
+        # If no detections, use the original frame
+        annotated_frame = frame.copy()
+
+    # Write the annotated frame to the output video
+    out.write(annotated_frame)
+
+    frame_count += 1
+    print(f"Processed frame {frame_count}", end='\r')
+
+# Release resources
+cap.release()
+out.release()
+print("\nInference complete. Video saved at", output_video_path)

models/waldo30/source.txt

@@ -0,0 +1 @@
+https://huggingface.co/StephanST/WALDO30