thtrieu/yolotf
Yolotf
yolotf is an object detection framework that provides tools for converting Darknet model configurations and weights into TensorFlow graphs. It includes a TensorFlow model trainer for training new detection models or fine-tuning existing weights using custom datasets.
The project features a mobile model exporter that serializes graph definitions and metadata into protobuf files for deployment on mobile devices.
The framework supports object detection inference on images and video to identify objects and export bounding box coordinates. It manages model state through weight-mapping translation and checkpoint-based training to allow for the restoration of weights and optimizer states.
Features
- Model Graph Conversions - Translates Darknet network configurations into TensorFlow computational graphs.
- Object Detection - Provides a framework for running inference on images and video to identify objects.
- Video Stream Detections - Processes video streams to identify objects and export annotated output video with bounding boxes.
- Detection Model Training - Provides capabilities for training or fine-tuning detection models on custom datasets using TensorFlow.
- Object Detection - Provides a Python interface for running object detection inference on image arrays.
- Weight Parameter Mapping - Maps binary weight tensors from Darknet files to TensorFlow variables to preserve model state.
- Cross-Framework Model Conversion - Translates Darknet model configurations and weights into TensorFlow graphs.
- Framework Format Converters - Translates model configurations and weights from Darknet format into TensorFlow graphs.
- Image Processing Pipelines - Ships an image processing pipeline that normalizes raw buffers for neural network inference.
- Training Checkpointing - Implements training checkpointing to save and restore model weights and optimizer states for fault tolerance.
- Mobile Model Serialization - Serializes trained model graphs and metadata into protobuf files for mobile deployment.
- Mobile Model Deployment - Exports model graphs and metadata as protobuf files for deployment on mobile devices.
- Model Export Utilities - Saves graph definitions and metadata to protobuf files for mobile deployment.
- Model Implementations - Real-time object detection optimized for mobile devices.
