Step1 : Change CIFAR-10 images into TFRecord
Step2 : Feeding TFRecord into Neural Network
Step3 : Get image recognizing result
If you want to change the amount of labels, you should change the "onehot" in the code.
test.v2：train_v2 contains 0~1 two classes, and each has about 100 images
test.v3：train_v3 contains 0~9 ten classes, and each has 100 images
test.v4：train_v4 contains 0~1 two classes, and each has 1000 images
test.v5：try to use another NN architecture Now want to shuffle images in advance
--> try to use high resolution images to solve the problem of reshaping
In conclusion, the result is that the image dataset's rewolution is too low, leading to reshape error Solution: replace with image for higher res olution
train.tfrecords_v6：try to collect 2 classes images with size biger than 70 KB
test.v6 + train_v6：try to classify 10 classes with AlexNet Use higher resolution images instead of all images to resolve reshape problem