---
library_name: transformers
tags: []
---

## Original result
```
IoU metric: bbox
 Average Precision  (AP) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.000
 Average Precision  (AP) @[ IoU=0.50      | area=   all | maxDets=100 ] = 0.000
 Average Precision  (AP) @[ IoU=0.75      | area=   all | maxDets=100 ] = 0.000
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = -1.000
 Average Precision  (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.000
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.000
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=  1 ] = 0.000
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets= 10 ] = 0.001
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.002
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = -1.000
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.000
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.002
```

## After training result
```
IoU metric: bbox
 Average Precision  (AP) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.028
 Average Precision  (AP) @[ IoU=0.50      | area=   all | maxDets=100 ] = 0.075
 Average Precision  (AP) @[ IoU=0.75      | area=   all | maxDets=100 ] = 0.021
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = -1.000
 Average Precision  (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.000
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.029
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=  1 ] = 0.089
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets= 10 ] = 0.152
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.166
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = -1.000
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.000
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.168
```

## Config
- dataset: NIH
- original model: hustvl/yolos-tiny
- lr: 0.0001
- dropout_rate: 0.1
- weight_decay: 1.0
- max_epochs: 30
- train samples: 885

## Logging
### Training process
```
{'validation_loss': tensor(7.2510, device='cuda:0'), 'validation_loss_ce': tensor(2.7062, device='cuda:0'), 'validation_loss_bbox': tensor(0.5285, device='cuda:0'), 'validation_loss_giou': tensor(0.9513, device='cuda:0'), 'validation_cardinality_error': tensor(99., device='cuda:0')}
{'training_loss': tensor(2.0733, device='cuda:0'), 'train_loss_ce': tensor(0.4772, device='cuda:0'), 'train_loss_bbox': tensor(0.1542, device='cuda:0'), 'train_loss_giou': tensor(0.4126, device='cuda:0'), 'train_cardinality_error': tensor(1., device='cuda:0'), 'validation_loss': tensor(2.3407, device='cuda:0'), 'validation_loss_ce': tensor(0.4447, device='cuda:0'), 'validation_loss_bbox': tensor(0.1600, device='cuda:0'), 'validation_loss_giou': tensor(0.5479, device='cuda:0'), 'validation_cardinality_error': tensor(1., device='cuda:0')}
{'training_loss': tensor(2.3671, device='cuda:0'), 'train_loss_ce': tensor(0.3991, device='cuda:0'), 'train_loss_bbox': tensor(0.1568, device='cuda:0'), 'train_loss_giou': tensor(0.5919, device='cuda:0'), 'train_cardinality_error': tensor(1., device='cuda:0'), 'validation_loss': tensor(2.3726, device='cuda:0'), 'validation_loss_ce': tensor(0.4201, device='cuda:0'), 'validation_loss_bbox': tensor(0.1725, device='cuda:0'), 'validation_loss_giou': tensor(0.5451, device='cuda:0'), 'validation_cardinality_error': tensor(1., device='cuda:0')}
{'training_loss': tensor(2.8530, device='cuda:0'), 'train_loss_ce': tensor(0.4628, device='cuda:0'), 'train_loss_bbox': tensor(0.2022, device='cuda:0'), 'train_loss_giou': tensor(0.6896, device='cuda:0'), 'train_cardinality_error': tensor(1., device='cuda:0'), 'validation_loss': tensor(2.1740, device='cuda:0'), 'validation_loss_ce': tensor(0.4275, device='cuda:0'), 'validation_loss_bbox': tensor(0.1443, device='cuda:0'), 'validation_loss_giou': tensor(0.5126, device='cuda:0'), 'validation_cardinality_error': tensor(1., device='cuda:0')}
{'training_loss': tensor(2.4427, device='cuda:0'), 'train_loss_ce': tensor(0.4072, device='cuda:0'), 'train_loss_bbox': tensor(0.1748, device='cuda:0'), 'train_loss_giou': tensor(0.5806, device='cuda:0'), 'train_cardinality_error': tensor(1., device='cuda:0'), 'validation_loss': tensor(2.1416, device='cuda:0'), 'validation_loss_ce': tensor(0.4228, device='cuda:0'), 'validation_loss_bbox': tensor(0.1418, device='cuda:0'), 'validation_loss_giou': tensor(0.5049, device='cuda:0'), 'validation_cardinality_error': tensor(1., device='cuda:0')}
{'training_loss': tensor(2.2765, device='cuda:0'), 'train_loss_ce': tensor(0.4684, device='cuda:0'), 'train_loss_bbox': tensor(0.1437, device='cuda:0'), 'train_loss_giou': tensor(0.5449, device='cuda:0'), 'train_cardinality_error': tensor(1., device='cuda:0'), 'validation_loss': tensor(2.1062, device='cuda:0'), 'validation_loss_ce': tensor(0.4127, device='cuda:0'), 'validation_loss_bbox': tensor(0.1439, device='cuda:0'), 'validation_loss_giou': tensor(0.4869, device='cuda:0'), 'validation_cardinality_error': tensor(1., device='cuda:0')}
{'training_loss': tensor(1.9443, device='cuda:0'), 'train_loss_ce': tensor(0.3799, device='cuda:0'), 'train_loss_bbox': tensor(0.1032, device='cuda:0'), 'train_loss_giou': tensor(0.5243, device='cuda:0'), 'train_cardinality_error': tensor(1., device='cuda:0'), 'validation_loss': tensor(1.9993, device='cuda:0'), 'validation_loss_ce': tensor(0.4028, device='cuda:0'), 'validation_loss_bbox': tensor(0.1352, device='cuda:0'), 'validation_loss_giou': tensor(0.4602, device='cuda:0'), 'validation_cardinality_error': tensor(1., device='cuda:0')}
{'training_loss': tensor(1.7868, device='cuda:0'), 'train_loss_ce': tensor(0.3504, device='cuda:0'), 'train_loss_bbox': tensor(0.1213, device='cuda:0'), 'train_loss_giou': tensor(0.4149, device='cuda:0'), 'train_cardinality_error': tensor(1., device='cuda:0'), 'validation_loss': tensor(2.0064, device='cuda:0'), 'validation_loss_ce': tensor(0.3905, device='cuda:0'), 'validation_loss_bbox': tensor(0.1337, device='cuda:0'), 'validation_loss_giou': tensor(0.4736, device='cuda:0'), 'validation_cardinality_error': tensor(1., device='cuda:0')}
{'training_loss': tensor(1.8019, device='cuda:0'), 'train_loss_ce': tensor(0.3278, device='cuda:0'), 'train_loss_bbox': tensor(0.0954, device='cuda:0'), 'train_loss_giou': tensor(0.4986, device='cuda:0'), 'train_cardinality_error': tensor(1., device='cuda:0'), 'validation_loss': tensor(2.0142, device='cuda:0'), 'validation_loss_ce': tensor(0.3723, device='cuda:0'), 'validation_loss_bbox': tensor(0.1391, device='cuda:0'), 'validation_loss_giou': tensor(0.4733, device='cuda:0'), 'validation_cardinality_error': tensor(1., device='cuda:0')}
{'training_loss': tensor(1.7681, device='cuda:0'), 'train_loss_ce': tensor(0.2725, device='cuda:0'), 'train_loss_bbox': tensor(0.1038, device='cuda:0'), 'train_loss_giou': tensor(0.4883, device='cuda:0'), 'train_cardinality_error': tensor(1., device='cuda:0'), 'validation_loss': tensor(2.0256, device='cuda:0'), 'validation_loss_ce': tensor(0.3890, device='cuda:0'), 'validation_loss_bbox': tensor(0.1334, device='cuda:0'), 'validation_loss_giou': tensor(0.4848, device='cuda:0'), 'validation_cardinality_error': tensor(1., device='cuda:0')}
{'training_loss': tensor(2.7969, device='cuda:0'), 'train_loss_ce': tensor(0.3935, device='cuda:0'), 'train_loss_bbox': tensor(0.1829, device='cuda:0'), 'train_loss_giou': tensor(0.7444, device='cuda:0'), 'train_cardinality_error': tensor(1., device='cuda:0'), 'validation_loss': tensor(2.0852, device='cuda:0'), 'validation_loss_ce': tensor(0.3799, device='cuda:0'), 'validation_loss_bbox': tensor(0.1377, device='cuda:0'), 'validation_loss_giou': tensor(0.5084, device='cuda:0'), 'validation_cardinality_error': tensor(1., device='cuda:0')}
{'training_loss': tensor(2.2780, device='cuda:0'), 'train_loss_ce': tensor(0.3727, device='cuda:0'), 'train_loss_bbox': tensor(0.1636, device='cuda:0'), 'train_loss_giou': tensor(0.5436, device='cuda:0'), 'train_cardinality_error': tensor(1., device='cuda:0'), 'validation_loss': tensor(1.8908, device='cuda:0'), 'validation_loss_ce': tensor(0.3772, device='cuda:0'), 'validation_loss_bbox': tensor(0.1258, device='cuda:0'), 'validation_loss_giou': tensor(0.4424, device='cuda:0'), 'validation_cardinality_error': tensor(1., device='cuda:0')}
{'training_loss': tensor(2.1853, device='cuda:0'), 'train_loss_ce': tensor(0.4098, device='cuda:0'), 'train_loss_bbox': tensor(0.1223, device='cuda:0'), 'train_loss_giou': tensor(0.5820, device='cuda:0'), 'train_cardinality_error': tensor(1., device='cuda:0'), 'validation_loss': tensor(1.8623, device='cuda:0'), 'validation_loss_ce': tensor(0.3716, device='cuda:0'), 'validation_loss_bbox': tensor(0.1160, device='cuda:0'), 'validation_loss_giou': tensor(0.4554, device='cuda:0'), 'validation_cardinality_error': tensor(1., device='cuda:0')}
{'training_loss': tensor(2.1250, device='cuda:0'), 'train_loss_ce': tensor(0.3339, device='cuda:0'), 'train_loss_bbox': tensor(0.1228, device='cuda:0'), 'train_loss_giou': tensor(0.5886, device='cuda:0'), 'train_cardinality_error': tensor(1., device='cuda:0'), 'validation_loss': tensor(1.8199, device='cuda:0'), 'validation_loss_ce': tensor(0.3661, device='cuda:0'), 'validation_loss_bbox': tensor(0.1161, device='cuda:0'), 'validation_loss_giou': tensor(0.4366, device='cuda:0'), 'validation_cardinality_error': tensor(1., device='cuda:0')}
{'training_loss': tensor(1.8382, device='cuda:0'), 'train_loss_ce': tensor(0.3676, device='cuda:0'), 'train_loss_bbox': tensor(0.1031, device='cuda:0'), 'train_loss_giou': tensor(0.4776, device='cuda:0'), 'train_cardinality_error': tensor(1., device='cuda:0'), 'validation_loss': tensor(1.9126, device='cuda:0'), 'validation_loss_ce': tensor(0.3584, device='cuda:0'), 'validation_loss_bbox': tensor(0.1291, device='cuda:0'), 'validation_loss_giou': tensor(0.4544, device='cuda:0'), 'validation_cardinality_error': tensor(1., device='cuda:0')}
{'training_loss': tensor(2.2184, device='cuda:0'), 'train_loss_ce': tensor(0.3580, device='cuda:0'), 'train_loss_bbox': tensor(0.1687, device='cuda:0'), 'train_loss_giou': tensor(0.5085, device='cuda:0'), 'train_cardinality_error': tensor(1., device='cuda:0'), 'validation_loss': tensor(1.9112, device='cuda:0'), 'validation_loss_ce': tensor(0.3618, device='cuda:0'), 'validation_loss_bbox': tensor(0.1275, device='cuda:0'), 'validation_loss_giou': tensor(0.4559, device='cuda:0'), 'validation_cardinality_error': tensor(1., device='cuda:0')}
{'training_loss': tensor(1.4304, device='cuda:0'), 'train_loss_ce': tensor(0.2193, device='cuda:0'), 'train_loss_bbox': tensor(0.0824, device='cuda:0'), 'train_loss_giou': tensor(0.3997, device='cuda:0'), 'train_cardinality_error': tensor(1., device='cuda:0'), 'validation_loss': tensor(1.9278, device='cuda:0'), 'validation_loss_ce': tensor(0.3509, device='cuda:0'), 'validation_loss_bbox': tensor(0.1320, device='cuda:0'), 'validation_loss_giou': tensor(0.4585, device='cuda:0'), 'validation_cardinality_error': tensor(1., device='cuda:0')}
{'training_loss': tensor(1.8228, device='cuda:0'), 'train_loss_ce': tensor(0.4408, device='cuda:0'), 'train_loss_bbox': tensor(0.1186, device='cuda:0'), 'train_loss_giou': tensor(0.3946, device='cuda:0'), 'train_cardinality_error': tensor(1., device='cuda:0'), 'validation_loss': tensor(1.8695, device='cuda:0'), 'validation_loss_ce': tensor(0.3668, device='cuda:0'), 'validation_loss_bbox': tensor(0.1232, device='cuda:0'), 'validation_loss_giou': tensor(0.4434, device='cuda:0'), 'validation_cardinality_error': tensor(1., device='cuda:0')}
{'training_loss': tensor(1.4936, device='cuda:0'), 'train_loss_ce': tensor(0.2966, device='cuda:0'), 'train_loss_bbox': tensor(0.0936, device='cuda:0'), 'train_loss_giou': tensor(0.3645, device='cuda:0'), 'train_cardinality_error': tensor(1., device='cuda:0'), 'validation_loss': tensor(1.8164, device='cuda:0'), 'validation_loss_ce': tensor(0.3554, device='cuda:0'), 'validation_loss_bbox': tensor(0.1208, device='cuda:0'), 'validation_loss_giou': tensor(0.4286, device='cuda:0'), 'validation_cardinality_error': tensor(1., device='cuda:0')}
{'training_loss': tensor(1.5842, device='cuda:0'), 'train_loss_ce': tensor(0.3072, device='cuda:0'), 'train_loss_bbox': tensor(0.0867, device='cuda:0'), 'train_loss_giou': tensor(0.4217, device='cuda:0'), 'train_cardinality_error': tensor(1., device='cuda:0'), 'validation_loss': tensor(1.9658, device='cuda:0'), 'validation_loss_ce': tensor(0.3572, device='cuda:0'), 'validation_loss_bbox': tensor(0.1300, device='cuda:0'), 'validation_loss_giou': tensor(0.4792, device='cuda:0'), 'validation_cardinality_error': tensor(1., device='cuda:0')}
{'training_loss': tensor(1.4302, device='cuda:0'), 'train_loss_ce': tensor(0.3983, device='cuda:0'), 'train_loss_bbox': tensor(0.0977, device='cuda:0'), 'train_loss_giou': tensor(0.2716, device='cuda:0'), 'train_cardinality_error': tensor(1., device='cuda:0'), 'validation_loss': tensor(1.9079, device='cuda:0'), 'validation_loss_ce': tensor(0.3678, device='cuda:0'), 'validation_loss_bbox': tensor(0.1266, device='cuda:0'), 'validation_loss_giou': tensor(0.4535, device='cuda:0'), 'validation_cardinality_error': tensor(0.9596, device='cuda:0')}
{'training_loss': tensor(1.8392, device='cuda:0'), 'train_loss_ce': tensor(0.3940, device='cuda:0'), 'train_loss_bbox': tensor(0.1368, device='cuda:0'), 'train_loss_giou': tensor(0.3806, device='cuda:0'), 'train_cardinality_error': tensor(1., device='cuda:0'), 'validation_loss': tensor(2.1207, device='cuda:0'), 'validation_loss_ce': tensor(0.3700, device='cuda:0'), 'validation_loss_bbox': tensor(0.1498, device='cuda:0'), 'validation_loss_giou': tensor(0.5008, device='cuda:0'), 'validation_cardinality_error': tensor(1., device='cuda:0')}
{'training_loss': tensor(2.0952, device='cuda:0'), 'train_loss_ce': tensor(0.3566, device='cuda:0'), 'train_loss_bbox': tensor(0.1601, device='cuda:0'), 'train_loss_giou': tensor(0.4692, device='cuda:0'), 'train_cardinality_error': tensor(1., device='cuda:0'), 'validation_loss': tensor(2.0796, device='cuda:0'), 'validation_loss_ce': tensor(0.4127, device='cuda:0'), 'validation_loss_bbox': tensor(0.1413, device='cuda:0'), 'validation_loss_giou': tensor(0.4801, device='cuda:0'), 'validation_cardinality_error': tensor(1., device='cuda:0')}
{'training_loss': tensor(1.7375, device='cuda:0'), 'train_loss_ce': tensor(0.3531, device='cuda:0'), 'train_loss_bbox': tensor(0.1156, device='cuda:0'), 'train_loss_giou': tensor(0.4033, device='cuda:0'), 'train_cardinality_error': tensor(1., device='cuda:0'), 'validation_loss': tensor(1.9825, device='cuda:0'), 'validation_loss_ce': tensor(0.3733, device='cuda:0'), 'validation_loss_bbox': tensor(0.1305, device='cuda:0'), 'validation_loss_giou': tensor(0.4784, device='cuda:0'), 'validation_cardinality_error': tensor(1., device='cuda:0')}
{'training_loss': tensor(2.6773, device='cuda:0'), 'train_loss_ce': tensor(0.3795, device='cuda:0'), 'train_loss_bbox': tensor(0.2213, device='cuda:0'), 'train_loss_giou': tensor(0.5956, device='cuda:0'), 'train_cardinality_error': tensor(1., device='cuda:0'), 'validation_loss': tensor(2.0580, device='cuda:0'), 'validation_loss_ce': tensor(0.3698, device='cuda:0'), 'validation_loss_bbox': tensor(0.1406, device='cuda:0'), 'validation_loss_giou': tensor(0.4927, device='cuda:0'), 'validation_cardinality_error': tensor(1., device='cuda:0')}
{'training_loss': tensor(2.0451, device='cuda:0'), 'train_loss_ce': tensor(0.3619, device='cuda:0'), 'train_loss_bbox': tensor(0.1562, device='cuda:0'), 'train_loss_giou': tensor(0.4512, device='cuda:0'), 'train_cardinality_error': tensor(1., device='cuda:0'), 'validation_loss': tensor(2.0350, device='cuda:0'), 'validation_loss_ce': tensor(0.3855, device='cuda:0'), 'validation_loss_bbox': tensor(0.1382, device='cuda:0'), 'validation_loss_giou': tensor(0.4791, device='cuda:0'), 'validation_cardinality_error': tensor(1., device='cuda:0')}
{'training_loss': tensor(2.1357, device='cuda:0'), 'train_loss_ce': tensor(0.3901, device='cuda:0'), 'train_loss_bbox': tensor(0.1723, device='cuda:0'), 'train_loss_giou': tensor(0.4422, device='cuda:0'), 'train_cardinality_error': tensor(1., device='cuda:0'), 'validation_loss': tensor(1.9736, device='cuda:0'), 'validation_loss_ce': tensor(0.3603, device='cuda:0'), 'validation_loss_bbox': tensor(0.1357, device='cuda:0'), 'validation_loss_giou': tensor(0.4673, device='cuda:0'), 'validation_cardinality_error': tensor(0.8687, device='cuda:0')}
{'training_loss': tensor(1.5814, device='cuda:0'), 'train_loss_ce': tensor(0.3938, device='cuda:0'), 'train_loss_bbox': tensor(0.0935, device='cuda:0'), 'train_loss_giou': tensor(0.3600, device='cuda:0'), 'train_cardinality_error': tensor(1., device='cuda:0'), 'validation_loss': tensor(2.0607, device='cuda:0'), 'validation_loss_ce': tensor(0.3561, device='cuda:0'), 'validation_loss_bbox': tensor(0.1488, device='cuda:0'), 'validation_loss_giou': tensor(0.4804, device='cuda:0'), 'validation_cardinality_error': tensor(0.9394, device='cuda:0')}
{'training_loss': tensor(1.7642, device='cuda:0'), 'train_loss_ce': tensor(0.3312, device='cuda:0'), 'train_loss_bbox': tensor(0.0892, device='cuda:0'), 'train_loss_giou': tensor(0.4934, device='cuda:0'), 'train_cardinality_error': tensor(1., device='cuda:0'), 'validation_loss': tensor(1.9519, device='cuda:0'), 'validation_loss_ce': tensor(0.3478, device='cuda:0'), 'validation_loss_bbox': tensor(0.1330, device='cuda:0'), 'validation_loss_giou': tensor(0.4697, device='cuda:0'), 'validation_cardinality_error': tensor(0.9697, device='cuda:0')}
{'training_loss': tensor(1.3615, device='cuda:0'), 'train_loss_ce': tensor(0.3474, device='cuda:0'), 'train_loss_bbox': tensor(0.0976, device='cuda:0'), 'train_loss_giou': tensor(0.2630, device='cuda:0'), 'train_cardinality_error': tensor(1., device='cuda:0'), 'validation_loss': tensor(1.8283, device='cuda:0'), 'validation_loss_ce': tensor(0.3620, device='cuda:0'), 'validation_loss_bbox': tensor(0.1182, device='cuda:0'), 'validation_loss_giou': tensor(0.4376, device='cuda:0'), 'validation_cardinality_error': tensor(0.9596, device='cuda:0')}
{'training_loss': tensor(1.4507, device='cuda:0'), 'train_loss_ce': tensor(0.2499, device='cuda:0'), 'train_loss_bbox': tensor(0.0839, device='cuda:0'), 'train_loss_giou': tensor(0.3905, device='cuda:0'), 'train_cardinality_error': tensor(1., device='cuda:0'), 'validation_loss': tensor(2.0212, device='cuda:0'), 'validation_loss_ce': tensor(0.3600, device='cuda:0'), 'validation_loss_bbox': tensor(0.1371, device='cuda:0'), 'validation_loss_giou': tensor(0.4878, device='cuda:0'), 'validation_cardinality_error': tensor(1., device='cuda:0')}
```

## Examples
{'size': tensor([512, 512]), 'image_id': tensor([1]), 'class_labels': tensor([4]), 'boxes': tensor([[0.2622, 0.5729, 0.0847, 0.0773]]), 'area': tensor([1717.9431]), 'iscrowd': tensor([0]), 'orig_size': tensor([1024, 1024])}

![Example](./example.png)