TBrain_Orchid Species Identification and Classification Competition

This repository contains code and models for the TBrain competition, which focuses on using AI, machine learning, and deep learning to train high-accuracy orchid image recognition models.

Preparation and Process

1. Use YOLOv5 to extract multiple bounding boxes from each training image.
2. Train the model using the segmented training images.
3. Upon receiving public data and private data, first use YOLOv5 to extract a bounding box from each image.
4. Then use each trained model to predict the category.
5. Finally, use the three CSV files predicted by the three trained models to perform weighted averaging.

Code Description

• run_SPM.py & run_FH.py:
  Two different methods for training models and making predictions.
• data_loader.py:
  Read and organize all the provided official data.
• Image_Search_Crawler.py:
  Crawl images from the web using image search.
• transpose.py:
  Data augmentation methods.