Using the pre-trained ImageNet models and cyclical learning rates, we tried to classify the DeepSAT-6 dataset (https://csc.lsu.edu/~saikat/deepsat/) into 6 categories (barren land, trees, grassland, roads, buildings and water bodies). Due to the absence of occlusion by the cloud, we dropped the NIR channel of the data.
Using the pre-trained ImageNet models and cyclical learning rates, we tried to classify the DeepSAT-6 dataset (https://csc.lsu.edu/~saikat/deepsat/) into 6 categories (barren land, trees, grassland, roads, buildings and water bodies). Due to absence of occlusion by cloud we dropped the NIR-channel of the data.
Using the pre-trained ImageNet models and cyclical learning rates, we tried to classify the DeepSAT-6 dataset (https://csc.lsu.edu/~saikat/deepsat/) into 6 categories (barren land, trees, grassland, roads, buildings and water bodies). Due to the absence of occlusion by the cloud, we dropped the NIR channel of the data.