ResNet9 model weights trained on animal sound data (DCASE 2023 Task 5's training set)

Checkpoint for ResNet9 model weights trained on DCASE 2023 Task 5's training set using supervised contrastive learning (following "Regularized Contrastive Pre-training for Few-shot Bioacoustic Sound Detection" https://arxiv.org/pdf/2309.08971).
This model is used in this repository: https://github.com/ilyassmoummad/BioAcousticInference for annotating audio files of animal sounds using very few manual annotation (as little as one).
PS: the weights contain an additional keys for a linear layer, you can load the weights use load_state_dict() with the argument strict=False to avoid errors related to the architecture expecting a linear layer.