My Version of Tensorflows Object Detection API.
The Idea was to create a realtime capable object detection pipeline on various machines.
Plug and play, ready to use without deep previous knowledge.
The following work has been done based on the original API:
- Capturing frames of a Camera-Input using OpenCV in seperate thread to increase performance
- Calculate fps, print the current value to console in a given intervall aswell as the overall mean value at the end
- Allows Models to grow GPU memory allocation. (ssd_mobilenet_v11_coco needs 350 MB)
- Added Option for detection without visualization to increase performance
- Added optional automated model download from model-zoo if necessary
- Added a script to be able to create tfEvent-files for Tensorboard Graph visualizationt
- Gathered necessary files to be able to export new frozen Models based on trained Checkpoints
- Exported new frozen Model based on ssd_mobilenet_v1_coco with altered score_threshold for batch_non_max_suppression to increase perfomance
- Added a script to be able to create tfEvent-files for Tensorboard Graph visualization
- Results: Overall Performance Increase of up to 100% depending on the running system
- Optional: change INPUT PARAMS which can be passed to object_detection.detection
- For example: If you are not interested in visualization: set visualize to False.
- if you want to import the pre-trained frozen Model .pb file to Tensorboard to visualize the Graph,
run frozenmodel_to_tensorboard.py and follow the command line instructions
(opt: change MODEL_NAME inside if necessary) - run object_detection10.py or object_detection.py Scripts
- For JetsonTX2 run object_detectionjetson.py Script
- For change parameters such as video_input or fps config.yml Script
- Enjoy!
- Ubuntu 16.04
- Python 2.7
- Tensorflow 1.4
- OpenCV 3.3.1
- Dell Laptop with i7 and GeForce GTX 1050: 35fps | 45fps
- Nvidia Jetson Tx2: 8fps | 12 fps
Pull request and open an issue are very welcome! 👍