This is a repo to help support the paper entitled Enabling automated herbarium sheet image post-processing through neural network models for color reference chart detection.
A modified Keras implementation of Faster R-CNN for Herbaria CRC detection is included. The original Keras implementation, including its training/testing instructions may be found here: https://github.com/you359/Keras-FasterRCNN
You may find all of the trained models within this link: https://drive.google.com/drive/folders/14W3oqSEb1bpzsHk4GB3ABAwFZSY91IIH?usp=sharing
- Small CRC
- Region Proposal: version 17k, "mlp_histogram_v17k.hdf5"
- Classifier: version 6c, "discriminator-v6c.tflite"
- Large CRC
- Modified Faster R-CNN: "model_frcnn.hdf5"