Live Demo
Try the Gradio application here:
Hugging Face Space: https://huggingface.co/spaces/ashutosh-kedar/ECO-VISION
Upload an image and the model detects:
- Trash
- Bin
- Hand
It also visualizes the predicted bounding boxes and confidence scores.
rt-detr-v2-finetuned-trash-hand-bin-bbox
This model is a fine-tuned version of PekingU/rtdetr_v2_r50vd on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 9.1691
- Map: 0.45
- Map 50: 0.62
- Map 75: 0.54
- Map Small: 0.05
- Map Medium: 0.2
- Map Large: 0.47
- Mar 1: 0.51
- Mar 10: 0.68
- Mar 100: 0.73
- Mar Small: 0.2
- Mar Medium: 0.61
- Mar Large: 0.74
- Map Bin: 0.75
- Mar Bin: 0.88
- Map Hand: 0.55
- Mar Hand: 0.78
- Map Not Bin: 0.01
- Mar Not Bin: 0.7
- Map Not Hand: 0.18
- Mar Not Hand: 0.45
- Map Not Trash: 0.21
- Mar Not Trash: 0.61
- Map Trash: 0.69
- Mar Trash: 0.82
- Map Trash Arm: 0.78
- Mar Trash Arm: 0.87
Model description
More information needed
Intended uses & limitations
More information needed
Training and evaluation data
More information needed
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 0.0001
- train_batch_size: 16
- eval_batch_size: 16
- seed: 42
- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 0.05
- num_epochs: 10
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Map | Map 50 | Map 75 | Map Small | Map Medium | Map Large | Mar 1 | Mar 10 | Mar 100 | Mar Small | Mar Medium | Mar Large | Map Bin | Mar Bin | Map Hand | Mar Hand | Map Not Bin | Mar Not Bin | Map Not Hand | Mar Not Hand | Map Not Trash | Mar Not Trash | Map Trash | Mar Trash | Map Trash Arm | Mar Trash Arm |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 81.8644 | 1.0 | 50 | 17.9704 | 0.15 | 0.21 | 0.17 | 0.0 | 0.04 | 0.16 | 0.25 | 0.38 | 0.47 | 0.0 | 0.24 | 0.49 | 0.5 | 0.85 | 0.38 | 0.84 | 0.0 | 0.17 | 0.0 | 0.0 | 0.01 | 0.48 | 0.16 | 0.79 | 0.0 | 0.13 |
| 25.1417 | 2.0 | 100 | 10.8062 | 0.39 | 0.54 | 0.43 | 0.01 | 0.15 | 0.4 | 0.42 | 0.61 | 0.66 | 0.3 | 0.45 | 0.68 | 0.72 | 0.87 | 0.58 | 0.82 | 0.01 | 0.74 | 0.0 | 0.0 | 0.11 | 0.55 | 0.61 | 0.81 | 0.7 | 0.8 |
| 18.3690 | 3.0 | 150 | 9.5975 | 0.41 | 0.57 | 0.5 | 0.01 | 0.18 | 0.43 | 0.45 | 0.68 | 0.73 | 0.3 | 0.56 | 0.75 | 0.75 | 0.88 | 0.58 | 0.83 | 0.02 | 0.77 | 0.0 | 0.35 | 0.18 | 0.6 | 0.66 | 0.83 | 0.71 | 0.87 |
| 16.0441 | 4.0 | 200 | 9.2287 | 0.44 | 0.59 | 0.51 | 0.01 | 0.2 | 0.45 | 0.48 | 0.7 | 0.75 | 0.4 | 0.58 | 0.76 | 0.76 | 0.88 | 0.55 | 0.81 | 0.01 | 0.8 | 0.01 | 0.45 | 0.21 | 0.61 | 0.69 | 0.83 | 0.82 | 0.87 |
| 14.6344 | 5.0 | 250 | 9.1273 | 0.45 | 0.59 | 0.52 | 0.04 | 0.2 | 0.46 | 0.46 | 0.7 | 0.74 | 0.3 | 0.59 | 0.75 | 0.75 | 0.88 | 0.57 | 0.82 | 0.01 | 0.7 | 0.02 | 0.45 | 0.25 | 0.63 | 0.69 | 0.83 | 0.82 | 0.9 |
| 13.6042 | 6.0 | 300 | 9.0929 | 0.42 | 0.6 | 0.46 | 0.01 | 0.21 | 0.44 | 0.5 | 0.69 | 0.73 | 0.1 | 0.57 | 0.75 | 0.76 | 0.88 | 0.56 | 0.8 | 0.01 | 0.75 | 0.08 | 0.4 | 0.22 | 0.61 | 0.69 | 0.82 | 0.63 | 0.87 |
| 12.8196 | 7.0 | 350 | 9.1643 | 0.43 | 0.61 | 0.46 | 0.01 | 0.21 | 0.44 | 0.49 | 0.69 | 0.73 | 0.1 | 0.57 | 0.74 | 0.76 | 0.88 | 0.55 | 0.8 | 0.01 | 0.72 | 0.1 | 0.4 | 0.22 | 0.61 | 0.69 | 0.82 | 0.65 | 0.87 |
| 12.2277 | 8.0 | 400 | 9.0923 | 0.45 | 0.61 | 0.53 | 0.08 | 0.2 | 0.46 | 0.5 | 0.68 | 0.73 | 0.2 | 0.58 | 0.74 | 0.76 | 0.88 | 0.55 | 0.79 | 0.01 | 0.71 | 0.1 | 0.4 | 0.23 | 0.62 | 0.69 | 0.82 | 0.78 | 0.9 |
| 11.7523 | 9.0 | 450 | 9.1509 | 0.44 | 0.61 | 0.53 | 0.05 | 0.2 | 0.46 | 0.49 | 0.69 | 0.72 | 0.2 | 0.58 | 0.73 | 0.75 | 0.88 | 0.55 | 0.78 | 0.01 | 0.67 | 0.1 | 0.4 | 0.22 | 0.62 | 0.69 | 0.82 | 0.78 | 0.87 |
| 11.5078 | 10.0 | 500 | 9.1691 | 0.45 | 0.62 | 0.54 | 0.05 | 0.2 | 0.47 | 0.51 | 0.68 | 0.73 | 0.2 | 0.61 | 0.74 | 0.75 | 0.88 | 0.55 | 0.78 | 0.01 | 0.7 | 0.18 | 0.45 | 0.21 | 0.61 | 0.69 | 0.82 | 0.78 | 0.87 |
Framework versions
- Transformers 5.12.0
- Pytorch 2.11.0+cu128
- Datasets 4.0.0
- Tokenizers 0.22.2