Can't improve val_acc, it's stuck on 0.5-0.6 idk what to do, need it with mobilenetv2 tried transfer learning with unlocking layers still same results, i used tf.keras.applications.mobilenet_v2.preprocess_input, and other parameters but still the same, anyone maybe know a good way to achieve a good results on mobilenetv2? I'm working on full dataset.0 Continue reading...
