I need a python development that should do the following
should pull images to train on from a directory structure. Each directory contains pictures of one class - directory name should be class name
It should change a pretrainend model so that it is optimised for the recognition of the number of classes.
It should then save the model and weights with keras standard methods so they can be used by the identification-component
The identification-component should :
Be a webservice taking in a image and it should return top 5 propositions. (Classname and probability ) in Json.
23 freelancers are bidding on average €550 for this job
Image recognition and object detection with boundaries are my favorites. You can trust me. Still can't.? You can even interview me before handing over the project. Am cool with it too.