WIP: Reading PNG images as arrays for MNIST Prediction

[sakshamarora1]

For Issue #371

Create image.csv

cat > image.csv << EOF
key, image
four, image1.png
five, image2.png
three, image3.png
two, image4.png
EOF

After training on the MNIST dataset ,
Predict command :

dffml predict all \
-model tfdnnc \
-model-batchsize 1000 \
-model-hidden 30 50 25 \
-model-clstype int \
-model-predict label:int:1 \
-model-classifications $(seq 0 9) \
-model-features image:int:$((28 * 28)) \
-sources images=csv \
-source-filename image.csv  \
-source-loadfiles image  \
-log debug

[johnandersen777]

Can you add what you made as the comment here to the docs/usage/mnist.rst file? Also, I recently started trying to get the documentation to be more testable, if you look at docs/quickstart/model.rst (all the literalincludes), examples/quickstart/, and examples/test_quickstart.py (in test_shell) you'll see how I was able to get the quickstart docs to be testable. It would be super great if we could make the mnist example testable like this as well!

[johnandersen777]

Let's also add a test for CSVSource to make sure that the loading which was added there works.

[johnandersen777]

Given yash's feedback, let's make the entrypoint on this mnistpng for now. So that no one gets confused. Let's leave the directory name though as png, thanks! Looking great :)

[johnandersen777]

@sakshamarora1 Looking great! Sorry to hit you with a couple more things but I just got around to responding. Thank you!

[johnandersen777]

Almost there! :D Nice work!!

[johnandersen777]

It looks like we also need to add to CORE_PLUGINS

diff --git a/dffml/service/dev.py b/dffml/service/dev.py
index b88818c9..f8fa2650 100644
--- a/dffml/service/dev.py
+++ b/dffml/service/dev.py
@@ -45,6 +45,7 @@ EMAIL = config.get("user", "email", fallback="[email protected]")
 
 CORE_PLUGINS = [
     ("configloader", "yaml"),
+    ("configloader", "png"),
     ("model", "tensorflow"),
     ("model", "scratch"),
     ("model", "scikit"),