run4.sh

git clone https://gist.github.com/257eef5b0e2d9e2625a9eb812ca2226b.git akmtdfgen
mv data akmtdfgen
cd akmtdfgen
python classifier_from_little_data_script_dfgen.py

Using TensorFlow backend.
Some samples:

df_train:
      object_id                                imgpath  target                          orig label
1683       1361  /tmp/path/to/imgs/361/01/cat_1361.jpg       0  data/train/cats/cat.1361.jpg   cat
1419       1072  /tmp/path/to/imgs/072/01/cat_1072.jpg       0  data/train/cats/cat.1072.jpg   cat
1040       1744  /tmp/path/to/imgs/744/01/cat_1744.jpg       0  data/train/cats/cat.1744.jpg   cat
91         1763  /tmp/path/to/imgs/763/01/dog_1763.jpg       1  data/train/dogs/dog.1763.jpg   dog
1061       1769  /tmp/path/to/imgs/769/01/cat_1769.jpg       0  data/train/cats/cat.1769.jpg   cat
df_valid:

     object_id                                imgpath  target                               orig label
727       7740  /tmp/path/to/imgs/740/07/cat_7740.jpg       0  data/validation/cats/cat.7740.jpg   cat
709       7656  /tmp/path/to/imgs/656/07/cat_7656.jpg       0  data/validation/cats/cat.7656.jpg   cat
82        7668  /tmp/path/to/imgs/668/07/dog_7668.jpg       1  data/validation/dogs/dog.7668.jpg   dog
270       7737  /tmp/path/to/imgs/737/07/dog_7737.jpg       1  data/validation/dogs/dog.7737.jpg   dog
676       7460  /tmp/path/to/imgs/460/07/cat_7460.jpg       0  data/validation/cats/cat.7460.jpg   cat

Training set: 2000 images, 2 classes.
Validation set: 888 images, 2 classes.

Epoch 1/5
44s - loss: 0.6821 - acc: 0.5910 - val_loss: 0.6439 - val_acc: 0.6295
Epoch 2/5
45s - loss: 0.5963 - acc: 0.6740 - val_loss: 0.7829 - val_acc: 0.5511
Epoch 3/5
44s - loss: 0.5294 - acc: 0.7505 - val_loss: 0.5699 - val_acc: 0.7102
Epoch 4/5
44s - loss: 0.4670 - acc: 0.7920 - val_loss: 0.6331 - val_acc: 0.6864
Epoch 5/5
45s - loss: 0.3846 - acc: 0.8440 - val_loss: 0.6407 - val_acc: 0.7034