Create RetinaNet object detection guide #2069
Conversation
sachinprasadhs
requested review from
fchollet,
MarkDaoust and
pcoet
as code owners
sachinprasadhs
requested review from
mattdangerw and
divyashreepathihalli
and removed request for
MarkDaoust and
pcoet
There was a problem hiding this comment.
Thanks for the PR @sachinprasadhs - since this is a guide - we should have more detailed explanations that would help users understand our framework better. Lets add more of that.
Thanks!
There was a problem hiding this comment.
Thanks! All minor comments. Also heads up I added keras-team/keras-hub#2219 for coco id - name mappings. Whenever it is released we can use it.
|
/gemini review |
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
| cache_dir=data_dir, | ||
| extract=True, | ||
| ) | ||
| data_dir = os.path.join(get_data, extracted_dir) |
There was a problem hiding this comment.
The keras.utils.get_file function returns the path to the downloaded file, not the directory where files are extracted. Use os.path.dirname to get the directory of the downloaded file.
| data_dir = os.path.join(get_data, extracted_dir) | |
| data_dir = os.path.join(os.path.dirname(get_data), extracted_dir) |
| year="2007", | ||
| split="trainval", | ||
| data_dir="./", | ||
| voc_url=VOC_2007_URL, | ||
| ): | ||
| extracted_dir = os.path.join("VOCdevkit", f"VOC{year}") |
There was a problem hiding this comment.
The image_file_paths variable is defined but not used. Remove it to avoid confusion.
| year="2007", | |
| split="trainval", | |
| data_dir="./", | |
| voc_url=VOC_2007_URL, | |
| ): | |
| extracted_dir = os.path.join("VOCdevkit", f"VOC{year}") | |
| # Parallel process all the images. | |
| annotation_file_paths = tf.io.gfile.glob( | |
| os.path.join(data_dir, "Annotations", "*.xml") | |
| ) |
| def decode_tfds(record): | ||
| """Decodes a standard TFDS object detection record. | ||
|
|
||
| Args: | ||
| record: A dictionary representing a single TFDS record. | ||
|
|
||
| Returns: | ||
| A dictionary with "images" and "bounding_boxes". | ||
| """ | ||
| image = record["image"] | ||
| image_shape = tf.shape(image) | ||
| height, width = image_shape[0], image_shape[1] | ||
| boxes = keras.utils.bounding_boxes.convert_format( | ||
| record["objects"]["bbox"], | ||
| source="rel_yxyx", | ||
| target=bbox_format, | ||
| height=height, | ||
| width=width, | ||
| ) | ||
| labels = record["objects"]["label"] | ||
|
|
||
| bounding_boxes = {"boxes": boxes, "labels": labels} | ||
|
|
||
| return {"images": image, "bounding_boxes": bounding_boxes} |
There was a problem hiding this comment.
The function decode_tfds is defined but never used. Remove it to improve maintainability.
| def preprocess_tfds(ds): | ||
| """Preprocesses a TFDS dataset for object detection. | ||
|
|
||
| Args: | ||
| ds: The TFDS dataset. | ||
| resizing: A resizing function. | ||
| max_box_layer: A max box processing function. | ||
| batch_size: The batch size. | ||
|
|
||
| Returns: | ||
| A preprocessed TFDS dataset. | ||
| """ | ||
| ds = ds.map(resizing, num_parallel_calls=tf.data.AUTOTUNE) | ||
| ds = ds.map(max_box_layer, num_parallel_calls=tf.data.AUTOTUNE) | ||
| ds = ds.batch(batch_size, drop_remainder=True) | ||
| return ds |
There was a problem hiding this comment.
The resizing and max_box_layer functions, and batch_size variable are used from the global scope, but are documented as arguments. Explicitly pass them as arguments to increase reusability.
| def preprocess_tfds(ds): | |
| """Preprocesses a TFDS dataset for object detection. | |
| Args: | |
| ds: The TFDS dataset. | |
| resizing: A resizing function. | |
| max_box_layer: A max box processing function. | |
| batch_size: The batch size. | |
| Returns: | |
| A preprocessed TFDS dataset. | |
| """ | |
| ds = ds.map(resizing, num_parallel_calls=tf.data.AUTOTUNE) | |
| ds = ds.map(max_box_layer, num_parallel_calls=tf.data.AUTOTUNE) | |
| ds = ds.batch(batch_size, drop_remainder=True) | |
| return ds | |
| def preprocess_tfds(ds, resizing, max_box_layer, batch_size): | |
| """Preprocesses a TFDS dataset for object detection. | |
| Args: | |
| ds: The TFDS dataset. | |
| resizing: A resizing function. | |
| max_box_layer: A max box processing function. | |
| batch_size: The batch size. | |
| Returns: | |
| A preprocessed TFDS dataset. | |
| """ | |
| ds = ds.map(resizing, num_parallel_calls=tf.data.AUTOTUNE) | |
| ds = ds.map(max_box_layer, num_parallel_calls=tf.data.AUTOTUNE) | |
| ds = ds.batch(batch_size, drop_remainder=True) | |
| return ds |
| update_freq=1, | ||
| ), | ||
| keras.callbacks.ModelCheckpoint( | ||
| ckpt_path + "/{epoch:04d}-{val_loss:.2f}.weights.h5", |
There was a problem hiding this comment.
Use os.path.join for constructing file paths to ensure platform independence.
| ckpt_path + "/{epoch:04d}-{val_loss:.2f}.weights.h5", | |
| os.path.join(ckpt_path, "{epoch:04d}-{val_loss:.2f}.weights.h5"), |
| keras-nlp 0.19.0 requires keras-hub==0.19.0, but you have keras-hub 0.20.0 which is incompatible. | ||
|
|
||
| WARNING: All log messages before absl::InitializeLog() is called are written to STDERR | ||
| E0000 00:00:1746815719.896182 8973 cuda_dnn.cc:8310] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered | ||
| E0000 00:00:1746815719.902635 8973 cuda_blas.cc:1418] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered |
There was a problem hiding this comment.
The output block contains a package incompatibility warning and some CUDA/cuDNN errors. Either resolve the dependency conflict, explain why these messages appear and if they can be safely ignored, or remove this output block from the guide to avoid confusion.
Summary:
This Pull Request introduces comprehensive documentation detailing how to use an object detection model from Keras Hub, using RetinaNet as an example.
This tutorial includes: