bird-feeder/BirdFSD-YOLOv5: BirdFSD-YOLOv5-v1.0.0-alpha.3

Training start time: 2022-05-03T01:33:35 Training duration: 06:58:12 W&B run URL: https://wandb.ai/biodiv/train/runs/sunuky9t W&B run ID: sunuky9t W&B run name: ruby-totem-47 W&B run path: biodiv/train/sunuky9t Number of classes: 21 Dataset name: dataset-YOLO-05-02-2022_20.00.33.tar

<details> <summary>System hardware</summary> <table> <tr> <td>CPU count <td>GPU count <td>GPU type <td>NVIDIA driver version <tr> <td>40 <td>1 <td>Tesla V100-PCIE-32GB <td>470.103.01 </table> </details><details> <summary>Base ML framework</summary> <table> <tr> <td>Python <td>CUDA <td>Torch <td>Torchvision <tr> <td>3.8.13 <td>10.2 <td>1.11.0 <td>0.12.0 </table> </details> <details> <summary>Classes</summary> ```YAML - Eastern Bluebird - Song Sparrow - Carolina Chickadee - Brown-headed Cowbird - Northern Cardinal - Mourning Dove - Chipping Sparrow - Spotted/Eastern Towhee (Rufous-sided Towhee) - Brown-headed Nuthatch - White-throated Sparrow - Carolina Wren - House Finch - Hermit Thrush - American Goldfinch - American Robin - squirrels - Downy Woodpecker - Red-bellied Woodpecker - Dark-eyed Junco - Tufted Titmouse - Yellow-rumped Warbler ``` <img src="https://s3.aibird.me/public/hist.jpg" alt="classes-hist"> </details><details> <summary>Validation predictions</summary> <img src="https://s3.aibird.me/public/Validation_100_33039ad52746cb016570.jpg" alt="val-0"> <img src="https://s3.aibird.me/public/Validation_100_62809e8011eff6d65202.jpg" alt="val-1"> <img src="https://s3.aibird.me/public/Validation_100_7b40c6aca6e31429d952.jpg" alt="val-2"> <img src="https://s3.aibird.me/public/Validation_100_f085fe8e93cb85eed027.jpg" alt="val-3"> <img src="https://s3.aibird.me/public/Validation_100_f6e5743019eddad40963.jpg" alt="val-4"> <img src="https://s3.aibird.me/public/Validation_100_fc7be025a19cf4280d40.jpg" alt="val-5"> </details><details> <summary>Confusion matrix</summary> <img src="https://s3.aibird.me/public/Results_100_f7549cf9432823ffd63e.png" alt="cm" width="1280" height="720"> </details>

Model performance:

<table> <tr> <td>best/precision <td>best/recall <td>val/box_loss <td>val/cls_loss <td>val/obj_loss <td>F1-score<tr> <td>0.9498 <td>0.8554 <td>0.0128 <td>0.0036 <td>0.005 <td>0.9002</table>