[WIP] cornell multigrasp preprocessing + loss (CRITICAL BUG IS PRESENT, NOT READY) #510
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
cornell_grasp_dataset_writer.py tfrecord singlegrasp & multigrasp filenames consolidate visualize_example call
ahundt
commented
Jun 9, 2018
| coordinate = tf.transpose(tf.convert_to_tensor( | ||
| [coordinate[1], | ||
| coordinate[0], | ||
| 1] | ||
| )) | ||
| elif len(coordinate.get_shape().as_list()) > 1: |
There was a problem hiding this comment.
There is definitely a bug around here somewhere. The gripper angle labels aren't rotating correctly with respect to the image rotations.
ahundt
commented
Jun 9, 2018
| @@ -842,7 +842,7 @@ def train_k_fold(split_type=None, | |||
| return run_histories | |||
|
|
|||
|
|
|||
| def k_fold_run_histories_average(run_histories, save_filename=None, metrics='val_acc', verbose=1): | |||
| def k_fold_run_histories_average(run_histories, save_filename=None, metrics='val_binary_accuracy', verbose=1): | |||
There was a problem hiding this comment.
need to actually specify the various metrics where this is called
ahundt
commented
Jun 9, 2018
| @@ -729,7 +776,10 @@ def get_simplified_features(feature_map): | |||
| dataset = dataset.map( | |||
| map_func=get_simplified_features, | |||
| num_parallel_calls=num_parallel_calls) | |||
| dataset = dataset.batch(batch_size=batch_size) # Parse the record into tensors. | |||
| if dynamic_pad: | |||
| dataset = dataset.padded_batch(batch_size=batch_size) | |||
There was a problem hiding this comment.
This doesn't work yet, might need to specify every single param that must be padded in the dictionary, and there are quite a few...
see the huge dictionary of different parameters in the dictionary variable named features in the function parse_and_preprocess()
I need ideas for this.
ahundt
changed the title
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Work towards support for multi grasp preprocessing and a loss function that takes the lowest loss of all proposed.
There is a critical bug in this PR's image preprocessing code where the grasp position and orientation will not match objects present in the image after preprocessing. This bug can be viewed by running
python2 cornell_grasp_dataset_reader.py, checking that inmain()the valuecrop_to = 'image_contains_grasp_box_center'is enabled.TODO: