Multi-scene citrus detection based on multi-task deep learning network

Several examples of multi task citrus detection on test data set:

Install

Check INSTALL.md for installation instructions.

Prepare Data

mkdir -p data/annotations
mkdir -p data/train
mkdir -p data/val
mkdir -p data/test
ln -s /path_to_data/annotations
ln -s /path_to_data/train
ln -s /path_to_data/val
ln -s /path_to_data/test

Model weights

mkdir weights
#The pretrained models will be downloaded when running the program.

My training log and pre-trained models can be found here link or link(pw:arn3).

Evaluation

To evalute the model, put the corresponding weights file in the ./weights directory and run one of the following commands.

# The model will detect the images in data/test/images, and output the detection results to data/test/result
python Multitask-Test.py

Training

To train the model, put the corresponding weights file in the ./weights directory and run one of the following commands.

# Put the pre training model with coco into logs and the model will be trained from data/train
python Multitask-Train.py

Results

License

This code is released under the MIT license. See LICENSE for additional details.

Thanks to the Third Party Libs

Mask_RCNN
Tensorflow