This is a PyTorch Tutorial to Object Detection.
This is blatanly "borrowed" from sgrvinod's great tutorial
Basic knowledge of PyTorch, convolutional neural networks is assumed.
- Run command
git clone https://github.com/tranduchuy682/SSD.git
pip install -r requirements.txt
- Or run .sh file
- The dataset can be downloaded or cloned here
- Format
├── AllDatabase
| ├── BCCD Database
| ├── LISCDatabase
| | └── Ground Truth Segmentation
| | └── Main Dataset
| | └── More Dataset without Ground Truth/alll
| ├── RaabinDatabase
| | └── GrTh
| | └── TestA
| | └── TestB/Test-B
| | └── Train
| ├── bbox.csv
| ├── test_bbox.csv
| └── train_bbox.csv
├── train.py
├── utils.py
...
- Run
cd ssd
git clone https://github.com/tranduchuy682/AllDatabase.git
- Run command
python3 create_data_lists.py
- Backbones: There are 3 backbone have been used
- VGG16 backbone
- Resnet18 backbone
- We have modified our resnet18 backbone to get resnet18* - more suitable with our dataset
- MobileNetV3 backbone
- Config
backbone == ["resnet18","mobilenetv3","vgg16"]
epoch = 150 or any int value- Run command
python3 train.py resnet18 150
- Run command
python3 eval.py
- Run command
python3 detect.py
- Run command
python3 app.py
- Demo
