STT-tensorflow/tensorflow/lite/java/ovic/Winner_OSS_Template.md

Project Name

Authors

The 1st place winner of the 4th On-device Visual Intelligence Competition (OVIC) of Low-Power Computer Vision Challenge (LPCVC)

• Last name, First name (@GitHubUsername)
• Last name, First name (@GitHubUsername)
• Last name, First name (@GitHubUsername)

Description

The model submitted for the OVIC and full implementation code for training, evaluation, and inference

• OVIC track: Image Classification, Object Detection

Algorithm

Requirements

To install requirements:

pip install -r requirements.txt

Pre-trained Models

Model	Download	MD5 checksum
Model Name	Download Link (Size: KB)	MD5 checksum

The model tar file contains the followings:

• Trained model checkpoint
• Frozen trained model
• TensorFlow Lite model

Results

4th OVIC Public Ranked Leaderboard

Image Classification (from the Leaderboard)

Rank	Username	Latency	Accuracy on Classified	# Classified	Accuracy/Time	Metric	Reference Accuracy
1	Username	xx.x	0.xxxx	20000.0	xxx	0.xxxxx	0.xxxxx

• Metric: Accuracy improvement over the reference accuracy from the Pareto optimal curve
• Accuracy on Classified: The accuracy in [0, 1] computed based only on the images classified within the wall-time
• # Classified: The number of images classified within the wall-time
• Accuracy/Time: The accuracy divided by either the total inference time or the wall-time, whichever is longer
• Reference accuracy: The reference accuracy of models from the Pareto optimal curve that have the same latency as the submission

Object Detection

Rank	Username	Metric	Runtime	mAP over time	mAP of processed
1	Username	0.xxxxx	xxx.x	xxx	xxx

• Metric: COCO mAP computed on the entire minival dataset
• mAP over time: COCO mAP on the minival dataset divided by latency per image
• mAP of processed: COCO mAP computed only on the processed images

Dataset

Training

Please run this command line for training.

python3 ...

Evaluation

Please run this command line for evaluation.

python3 ...

References

License

This project is licensed under the terms of the Apache License 2.0.

Citation

If you want to cite this repository in your research paper, please use the following information.