[WIP] Running detection code with DLP

[thunterdb]

This is an experimental branch for running some of the latest object detection code from the tensorflow research repository. It will be broken up once stabilized

Summary of the changes:

• includes #110 . Thank you @tomasatdatabricks !
• update to the nix script, for full reproducibility
• TFImageTransformer can return multiple outputs

[codecov-io]

Codecov Report

Merging #112 into master will decrease coverage by 12.14%. The diff coverage is n/a.

@@             Coverage Diff             @@
##           master     #112       +/-   ##
===========================================
- Coverage   85.37%   73.22%   -12.15%     
===========================================
  Files          33       10       -23     
  Lines        1921      493     -1428     
  Branches       44       44               
===========================================
- Hits         1640      361     -1279     
+ Misses        281      132      -149

Impacted Files	Coverage Δ
python/sparkdl/udf/keras_image_model.py
...n/sparkdl/estimators/keras_image_file_estimator.py
python/sparkdl/graph/input.py
python/sparkdl/transformers/keras_tensor.py
python/sparkdl/param/converters.py
python/sparkdl/transformers/named_image.py
python/sparkdl/param/__init__.py
python/sparkdl/transformers/keras_image.py
python/sparkdl/graph/builder.py
python/sparkdl/utils/jvmapi.py
... and 11 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update aa7acdb...4262220. Read the comment docs.

[thunterdb]

Closing this PR which is superseded by #171