Update chip_inference Code to Handle Larger Image Sizes

[rym-oualha]

The current implementation of the chip_inference code is designed to work with image sizes similar to those used during the training phase. However, it is necessary to update the code to handle larger image sizes efficiently and reliably. This would expand the applicability of the model to real-world use cases where image dimensions may exceed the training set's constraints.