I’m developing a workflow to identify natural fibres in second-hand fabrics by analysing optical-microscope images and then automatically sorting those images with machine learning. The emphasis is on natural fibres only—cotton, wool, silk, linen and any other plant- or animal-based fibres I add later. Here’s what I need from you: You need to collect samples of fabrics and take optical microscopic images to use it further to segregate using machine learning • A complete image-classification pipeline, preferably in Python using libraries such as TensorFlow, Keras, PyTorch or similar. • Clear guidelines for capturing consistent microscope images (lighting, magnification, slide prep) so I can expand the data set over time. • Data preprocessing scripts that normalise, augment and split the images for training, validation and testing. • A trained model that accurately distinguishes cotton, wool, silk, linen and an “other natural fibres” class, delivered with saved weights and the code required to reproduce training. • Performance metrics (confusion matrix, precision/recall, overall accuracy) plus a brief report explaining where the model excels or needs future improvement. • An inference script or lightweight API so the classifier can be called from another application to return the predicted fibre type and probability score. I’m comfortable performing additional data annotation if you provide a repeatable template, but the core classification model and documentation are your responsibility. If you can boost reliability with feature-extraction techniques or transfer learning, feel free to do so—accuracy and maintainability are my top priorities. Deliver clean, well-commented code, all dependencies listed in a requirements file, and concise usage instructions so I can run everything on my local machine without guesswork.