# YOLOV3 for thyroid nodules detection
We trained Yolov3 on the data set of 1805 b-ultrasound images of thyroid nodules. And the weights are stored in ./pretrained_model .The test interface is offered here.
The forward process of Yolov3 will run on the GPU by default.
Python 3.6 or later with the following pip install -r requirements.txt
packages:
- numpy
- torch>=1.0
- torchvision
- tensorflow
- pillow
- tqdm
- opencv
-
The input image format must be 'jpg' or 'png'.
-
Store all images to be tested in the same directory</test_path>, and the default path is./test_data .
-
Create a directory </store_path> under the project directory to store the test results, and the default path is ./test_result .
Script: test_my.py
Optional args:
- image_path -path to test images
- store_path -path to test results
- img_size -size of each image dimension(no change recommended here)
- n_cpu -number of CPU threads to use during batch generation
- pretrained_weights -path to pretrained weights
- model_def -path to model definition file
- batch_size -size of each image batch
Run:
python test_my.py --image_path <test_path> --store_path <store_path>
You can got result images in output folder.