Instructions to use deni1000/yolo26-text-bubble-detection with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- ultralytics
How to use deni1000/yolo26-text-bubble-detection with ultralytics:
# Couldn't find a valid YOLO version tag. # Replace XX with the correct version. from ultralytics import YOLOvXX model = YOLOvXX.from_pretrained("deni1000/yolo26-text-bubble-detection") source = 'http://images.cocodataset.org/val2017/000000039769.jpg' model.predict(source=source, save=True) - Notebooks
- Google Colab
- Kaggle
File size: 2,359 Bytes
94992a5 87b52f8 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 | ---
library_name: ultralytics
tags:
- object-detection
- yolo
- manga
- comic
- text-detection
- bubble-detection
metrics:
- map
---
## Model Performance
| Metric | Final | Best |
|---|:---:|:---:|
| mAP@50-95 (B) | 0.8434 | 0.8446 |
| mAP@50 (B) | 0.9687 | 0.9865 |
| Precision (B) | 0.9843 | 0.9949 |
| Recall (B) | 0.9511 | 0.9651 |
### Training Curves
<table>
<tr>
<td align="center"><b>mAP@50-95</b></td>
<td align="center"><b>mAP@50</b></td>
</tr>
<tr>
<td><img src="https://cdn-uploads.huggingface.co/production/uploads/69ae9e7638302cb833f1c09e/HAMMBlZWiWGBgImimpJ_y.jpeg" width="800"></td>
<td><img src="https://cdn-uploads.huggingface.co/production/uploads/69ae9e7638302cb833f1c09e/Odx5ADGSQqghrnqJ7IpL9.jpeg" width="800"></td>
</tr>
<tr>
<td align="center"><b>Precision</b></td>
<td align="center"><b>Recall</b></td>
</tr>
<tr>
<td><img src="https://cdn-uploads.huggingface.co/production/uploads/69ae9e7638302cb833f1c09e/SItT-nCqDEGzSgAgd9P-C.jpeg" width="800"></td>
<td><img src="https://cdn-uploads.huggingface.co/production/uploads/69ae9e7638302cb833f1c09e/disIT1r8L8tURvcl806_y.jpeg" width="800"></td>
</tr>
</table>
## Training Hyperparameters
### General
| Parameter | Value |
|---|---|
| Base model | `yolo26s.pt` |
| Task | `detect` |
| Image size | `1280` |
| Epochs | `1200` |
| Batch size | `16` |
| Nominal batch size (nbs) | `64` |
| Device | `1,0` (2× GPU) |
| Workers | `8` |
| Pretrained | `true` |
| AMP | `true` |
| Cache | `true` |
| Seed | `0` |
### Optimizer
| Parameter | Value |
|---|---|
| Optimizer | `MuSGD` |
| lr0 | `0.01` |
| lrf | `0.005` |
| Cosine LR | `true` |
| Momentum | `0.937` |
| Weight decay | `5e-4` |
| Warmup epochs | `15` |
| Warmup momentum | `0.8` |
| Warmup bias lr | `0.1` |
| Patience (early stop) | `10000` |
### Loss
| Parameter | Value |
|---|---|
| box | `7.5` |
| cls | `0.5` |
| dfl | `1.5` |
### Augmentation
| Parameter | Value |
|---|---|
| Mosaic | `0.7` |
| Close mosaic | `30` |
| fliplr | `0.5` |
| flipud | `0.0` |
| translate | `0.1` |
| scale | `0.7` |
| shear | `5` |
| degrees | `0.0` |
| perspective | `3e-4` |
| hsv_h | `0.015` |
| hsv_s | `0.7` |
| hsv_v | `0.4` |
| copy_paste | `0.0` |
| mixup | `0.0` |
| cutmix | `0.0` |
| erasing | `0` |
### Inference / NMS
| Parameter | Value |
|---|---|
| IoU threshold | `0.7` |
| Max detections | `300` |
|