[WIP] Tf object detection as grasp prediction

costar_google_brainrobotdata/README.md

@@ -33,4 +33,121 @@ Color augmentation is also available:
 4. vrep should load and start the simulation
 5. make sure the folder holding `vrep_grasp.py` is on your PYTHONPATH
 6. cd to `~/src/costar_ws/src/costar_plan/costar_google_brainrobotdata/`, or wherever you put the repository
-7. run `python2 vrep_grasp.py`
+7. run `export CUDA_VISIBLE_DEVICES="" && python2 vrep_grasp.py`
+
+## Hyperparameter search
+
+
+### Google Brain Grasping Dataset
+
+To run the search execute the following command
+
+```
+export CUDA_VISIBLE_DEVICES="0" && python2 google_grasp_hyperopt.py --run_name single_prediction_all_transforms
+```
+
+Generating a hyperparameter search results summary for google brain grasping dataset classification:
+
+```
+python hyperopt_rank.py --log_dir hyperopt_logs_google_brain_classification --sort_by val_acc
+```
+
+### Cornell Dataset
+
+#### Regression
+```
+
+```
+
+The following commands will generate a hyperopt summary.
+
+Generating a hyperparameter search results summary for cornell regression
+
+```
+python hyperopt_rank.py --log_dir hyperopt_logs_cornell_regression --sort_by val_grasp_jaccard
+```
+
+#### Classification
+
+```
+export CUDA_VISIBLE_DEVICES="0" && python cornell_hyperopt.py --log_dir hyperopt_logs_cornell_classification
+```
+
+
+Generating a hyperparameter search results summary for cornell classification
+
+```
+python hyperopt_rank.py --log_dir hyperopt_logs_cornell_classification --sort_by val_binary_accuracy
+```
+
+
+## TF Object Detection API
+
+
+1. follow the installation instructions for tensorflow_models.sh
+in github.com/ahundt/robotics_setup.
+2. That will install the [tf object detection
+API](https://github.com/tensorflow/models/blob/master/research/object_detection)
+3. Perform cornell dataset tfrecord generation for use in object detection
+
+```
+python2 cornell_grasp_dataset_writer --write True --padding False --angle_classes True --include_negative_examples False --angle_classes_zero_is_background_class True
+```
+
+4. Run `python tf_object_detection.py`, this will download models from the [tf object detection model zoo](https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/detection_model_zoo.md).
+
+Model files will be downloaded to
+
+- `~/.keras/datasets/faster_rcnn_inception_resnet_v2_atrous_coco_2018_01_28/`
+- `~/.keras/datasets/ssd_mobilenet_v2_coco_2018_03_29/`
+- `~/.keras/datasets/faster_rcnn_nas_coco_2018_01_28/`
+
+Next step is to set up configuration files for a training run.
+
+Some reference commands might be at https://github.com/tensorflow/models/issues/3909
+
+TODO(ahundt) how to specify multiple tfrecord files?
+(the next batch of text is from [train.py](https://github.com/tensorflow/models/blob/master/research/object_detection/train.py))
+
+The executable `research/object_detection/train.py` is used to train DetectionModels. There are two ways of
+configuring the training job:
+1) A single pipeline_pb2.TrainEvalPipelineConfig configuration file
+can be specified by --pipeline_config_path.
+Example usage:
+    ./train \
+        --logtostderr \
+        --train_dir=path/to/train_dir \
+        --pipeline_config_path=pipeline_config.pbtxt
+2) Three configuration files can be provided: a model_pb2.DetectionModel
+configuration file to define what type of DetectionModel is being trained, an
+input_reader_pb2.InputReader file to specify what training data will be used and
+a train_pb2.TrainConfig file to configure training parameters.
+Example usage:
+```
+    ./train \
+        --logtostderr \
+        --train_dir=path/to/train_dir \
+        --model_config_path=model_config.pbtxt \
+        --train_config_path=train_config.pbtxt \
+        --input_config_path=train_input_config.pbtxt
+```
+
+Here are two example commands to run one after the other from [github issue #1854](https://github.com/tensorflow/models/issues/1854) train with:
+
+```
+python object_detection/train.py \
+    --logtostderr \
+    --pipeline_config_path=~/.virtualenvs/Project/lib/python2.7/site-packages/tensorflow/models/object_detection/samples/configs/faster_rcnn_resnet101_pets_learn.config \
+    --train_dir=~/.virtualenvs/Project/models/model/train
+```
+
+Then run the following to evaluate:
+
+```
+python object_detection/eval.py \
+    --logtostderr \
+    --pipeline_config_path=~/.virtualenvs/Project/lib/python2.7/site-packages/tensorflow/models/object_detection/samples/configs/faster_rcnn_resnet101_pets_learn.config \
+    --checkpoint_dir=~/.virtualenvs/Project/models/model/train \
+    --eval_dir=~/.virtualenvs/Project/models/model/eval
+
+```