This repository provides a robust framework for recognizing various Russian dishes using the Faster R-CNN deep learning architecture. This project aims to assist in the automatic identification and categorization of Russian cuisine in images, which can be applied in diverse domains such as food blogging, dietary tracking, and restaurant automation.
The Russian Food Recognition project leverages the Faster R-CNN and SSD models to detect and classify different types of Russian food in images. Faster R-CNN is a state-of-the-art object detection model that provides high accuracy and speed, making it suitable for real-time food recognition applications.
- High Accuracy: Utilizes Faster R-CNN and SSD for precise food detection and classification.
- Extensive Dataset: Trained on a diverse dataset of Russian dishes.
- Scalability: Easily adaptable to include more food categories or different cuisines.
- Modular Design: Clear separation of data processing, model training, and evaluation modules.
You can vizualize this project in application
or run locally:
streamlit run app.py| Modules | Parameters |
|---|---|
| model.backbone.body.conv1.weight | 9408 |
| model.backbone.body.layer1.0.conv1.weight | 4096 |
| model.backbone.body.layer1.0.conv2.weight | 36864 |
| model.backbone.body.layer1.0.conv3.weight | 16384 |
| model.backbone.body.layer1.0.downsample.0.weight | 16384 |
| model.backbone.body.layer1.1.conv1.weight | 16384 |
| model.backbone.body.layer1.1.conv2.weight | 36864 |
| model.backbone.body.layer1.1.conv3.weight | 16384 |
| model.backbone.body.layer1.2.conv1.weight | 16384 |
| model.backbone.body.layer1.2.conv2.weight | 36864 |
| model.backbone.body.layer1.2.conv3.weight | 16384 |
| model.backbone.body.layer2.0.conv1.weight | 32768 |
| model.backbone.body.layer2.0.conv2.weight | 147456 |
| model.backbone.body.layer2.0.conv3.weight | 65536 |
| model.backbone.body.layer2.0.downsample.0.weight | 131072 |
| model.backbone.body.layer2.1.conv1.weight | 65536 |
| model.backbone.body.layer2.1.conv2.weight | 147456 |
| model.backbone.body.layer2.1.conv3.weight | 65536 |
| model.backbone.body.layer2.2.conv1.weight | 65536 |
| model.backbone.body.layer2.2.conv2.weight | 147456 |
| model.backbone.body.layer2.2.conv3.weight | 65536 |
| model.backbone.body.layer2.3.conv1.weight | 65536 |
| model.backbone.body.layer2.3.conv2.weight | 147456 |
| model.backbone.body.layer2.3.conv3.weight | 65536 |
| model.backbone.body.layer3.0.conv1.weight | 131072 |
| model.backbone.body.layer3.0.conv2.weight | 589824 |
| model.backbone.body.layer3.0.conv3.weight | 262144 |
| model.backbone.body.layer3.0.downsample.0.weight | 524288 |
| model.backbone.body.layer3.1.conv1.weight | 262144 |
| model.backbone.body.layer3.1.conv2.weight | 589824 |
| model.backbone.body.layer3.1.conv3.weight | 262144 |
| model.backbone.body.layer3.2.conv1.weight | 262144 |
| model.backbone.body.layer3.2.conv2.weight | 589824 |
| model.backbone.body.layer3.2.conv3.weight | 262144 |
| model.backbone.body.layer3.3.conv1.weight | 262144 |
| model.backbone.body.layer3.3.conv2.weight | 589824 |
| model.backbone.body.layer3.3.conv3.weight | 262144 |
| model.backbone.body.layer3.4.conv1.weight | 262144 |
| model.backbone.body.layer3.4.conv2.weight | 589824 |
| model.backbone.body.layer3.4.conv3.weight | 262144 |
| model.backbone.body.layer3.5.conv1.weight | 262144 |
| model.backbone.body.layer3.5.conv2.weight | 589824 |
| model.backbone.body.layer3.5.conv3.weight | 262144 |
| model.backbone.body.layer4.0.conv1.weight | 524288 |
| model.backbone.body.layer4.0.conv2.weight | 2359296 |
| model.backbone.body.layer4.0.conv3.weight | 1048576 |
| model.backbone.body.layer4.0.downsample.0.weight | 2097152 |
| model.backbone.body.layer4.1.conv1.weight | 1048576 |
| model.backbone.body.layer4.1.conv2.weight | 2359296 |
| model.backbone.body.layer4.1.conv3.weight | 1048576 |
| model.backbone.body.layer4.2.conv1.weight | 1048576 |
| model.backbone.body.layer4.2.conv2.weight | 2359296 |
| model.backbone.body.layer4.2.conv3.weight | 1048576 |
| model.backbone.fpn.inner_blocks.0.0.weight | 65536 |
| model.backbone.fpn.inner_blocks.0.0.bias | 256 |
| model.backbone.fpn.inner_blocks.1.0.weight | 131072 |
| model.backbone.fpn.inner_blocks.1.0.bias | 256 |
| model.backbone.fpn.inner_blocks.2.0.weight | 262144 |
| model.backbone.fpn.inner_blocks.2.0.bias | 256 |
| model.backbone.fpn.inner_blocks.3.0.weight | 524288 |
| model.backbone.fpn.inner_blocks.3.0.bias | 256 |
| model.backbone.fpn.layer_blocks.0.0.weight | 589824 |
| model.backbone.fpn.layer_blocks.0.0.bias | 256 |
| model.backbone.fpn.layer_blocks.1.0.weight | 589824 |
| model.backbone.fpn.layer_blocks.1.0.bias | 256 |
| model.backbone.fpn.layer_blocks.2.0.weight | 589824 |
| model.backbone.fpn.layer_blocks.2.0.bias | 256 |
| model.backbone.fpn.layer_blocks.3.0.weight | 589824 |
| model.backbone.fpn.layer_blocks.3.0.bias | 256 |
| model.rpn.head.conv.0.0.weight | 589824 |
| model.rpn.head.conv.0.0.bias | 256 |
| model.rpn.head.cls_logits.weight | 768 |
| model.rpn.head.cls_logits.bias | 3 |
| model.rpn.head.bbox_pred.weight | 3072 |
| model.rpn.head.bbox_pred.bias | 12 |
| model.roi_heads.box_head.fc6.weight | 12845056 |
| model.roi_heads.box_head.fc6.bias | 1024 |
| model.roi_heads.box_head.fc7.weight | 1048576 |
| model.roi_heads.box_head.fc7.bias | 1024 |
| model.roi_heads.box_predictor.cls_score.weight | 132096 |
| model.roi_heads.box_predictor.cls_score.bias | 129 |
| model.roi_heads.box_predictor.bbox_pred.weight | 528384 |
| model.roi_heads.box_predictor.bbox_pred.bias | 516 |
Total Trainable Params: 41 950 036
The dataset on HF and on Kaggle in compressed image format used for training consists of a variety of images representing different Russian dishes. Each image is annotated with bounding boxes and labels corresponding to the food items present.
- Make the project modular
- Finetuning YOLOv10, RT-DETR
- Configure TensorBoard, wandb
- Create application
- Rewrite file with annotations