This pipeline can be utilised for object detection usecases.
- Image and Bbox slicing along with annotations
- Image with and without ROI sampling along with annotations
- Annotation fromat conversion from Pascal to VOC
- Yolov5 model with all its variances, tensorboard and end to end logging
- Object detection inference using SAHI
- Model and Experiment versioning
- Dataset versioning (primary)
- REST Apis created
- Separate modules for each functionality
- Yaml files for configuration setup
- Adding experiment name in Yolov5 for training
- Quantised / Onnx model export feature
- Enable detect endpoint of Yolov5
- auto suggestion of region of overlap selection based on data for slicing
- Enable MLFlow logging
- Adding time for each requests
- Test cases creation
- Exception handling
- Dockerfile creation
- Continuous Integration testing
- label studio automated annotation Support
- integration with other object detection models
- Extending pipeline for segmentation use cases
- RetinaNet model integration with resnet/mobilenet backbones
- auto generation of prediction scripts
- auto generation of prediction Apis
- auto docker container generation and export
- logging and versioning of annotations
- Background processes Support
- Database Sqlite/Postgres Support
- Create a separate environment
- Install dependencies mentioned in requirements.txt
- Goto project folder and launch python environment
- local_start.bat
- Open the website followed by /docs
- Bijon Guha
Active