Train a Deep Learning Image Segmentation Model based on DeepLabV3

I need a CNN based image segmentation model including the pre-processing code, the training code, test code and inference code. The image segmentation model shall be trained on a dataset that I can provide. The basic principle of what the model shall do is depicted in '[login to view URL]' One example of a image pair from the taining set (input image + output segmentation) is given in '[login to view URL]' The training code shall also include data augmentation with following steps: - random image rotation - random image cropping - random color shifting - adding noise in different levels - adding blur to the image in different levels - adding a random background image (from a dataset like ImageNet dataset) regarding the last point: Please take a look at "[login to view URL]" The model architecture shall be based on an encoder - decoder network. The decoder backbone shall be a MobileNetV2 (or or possibly newer if it benefits the performance) and the decoder backbone shall be a DeepLabV3. The intended model architecture is depicted in '[login to view URL]' I understand the DeepLabV3 and MobileNetV2 requirements for the architecture are quite specifig and it might be possible that it takes extra effort to consider this. If you have understood that implication of the requirement, please include: "HAU38J" into you cover letter.