For Moth Project backup in BRCAS Laboratory for Ecology and Sociobiology Studies(ESS) Lab
-
Using fine-tuned object detection model (YOLOv4) for Lepidoptera to frame the exact location of specimen images
-
Obtain background removal images of Lepidoptera specimens (
Moth_thermal/remove_bg/)- Obtain a small number of rough images of the background removal images using "Unsupervised Image Segmentation by Backpropagation" model
- Supervised training of the model (UNet) from scratch, incremental increase, and accurate background removal images
-
Variational Sparse Coding (VSC) in Latent space based on variational autoencoders (VAEs) is used to improve feature interpretability for subsequent analysis of deep features(
Moth_thermal/vsc/)