Model summary

Here, two deep neural networks for primary grading of Korean commercial pig are proposed.

1. Back-fat thickness estimation network (BTENet), which simultaneously performs the back-fat area segmentation and thickness estimation.
2. Sex classification network (SCNet), which determines the sex classes of pig carcass.

Implementation

1. Preparing

Our models were implemented by tensorflow 2.3 in Python 3.8.6. Pre-trained weights for BTENet and SCNet can be downloaded from https://drive.google.com/drive/folders/1PRBpfRVALwiPbA6JYSB9V-jDl2Rr8FIJ?usp=sharing. important Both pre-trained weights must be placed at [CODE PATH]/weights for execution.

2. BTENet

2.1. Execution

python main_btenet.py --img [IMG] --out [OUT] --device [device]

1. [IMG]: image file path, which contain the head-side image of VCS-2000.
2. [OUT]: Path for saving the results.
3. [device]: GPU device number to use (default: 0).

example

python main_btenet.py --img data/head_img --out btenet_results --device 0

2.2 Output

1. bf.sol: A text file with a header line, and the one line per sample with 2 columnes. The first column is file name and another is predicted back-fat thickness (mm).
2. mask: A file path, which includes the predicted back-fat area mask.
3. paint: A file path, which includes visualized prediction results.

3. SCNet

2.1. Execution

python main_scnet.py --img [IMG] --out [OUT] --device [device]

1. [IMG]: image file path, which contain the hip-side image of VCS-2000.
2. [OUT]: Path for saving the results.
3. [device]: GPU device number to use (default: 0).

example

python main_scnet.py --img data/hip_img --out scnet_results --device 0

2.2 Output

1. pred.sol: A text file with a header line, and the one line per sample with 3 columnes. Each columns mean file name, predicted sex class, and class probability
2. paint: A file path, which includes visualized prediction results.