SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

[Proposal] Breaking change zero-shot-object-detection for improved consistency. (#20280)

Browse Source 
* [Proposal] Breaking change `zero-shot-object-detection` for improved
consistency.

This is a proposal to modify the output of `zero-shot-object-detection`
to provide better alignment with other pipelines.

The output is now strictly the same as `object-detection` whereas before
it would output lists of lists.

The name `candidate_labels` is used throughout for consistency with
other `zero-shot` pipelines.

The pipeline is changed to `ChunkPipeline` to support batching cleanly.

This removes all the lists and list of lists shenanigans, it's now a
matter of the base pipeline handling all this not this specific one.

**Breaking change**: It did remove complex calls potentials `pipe(images = [image1, image2],
text_queries=[candidates1, candidates2])` to support only
`pipe([{"image": image1, "candidate_labels": candidates1}, {"image": image2, "candidate_labels": candidates2}])`
when dealing with lists and/or datasets.
We could keep them, but it will add a lot of complexity to the code
base, since the pipeline is rather young, I'd rather break to keep the
code simpler, but we can revert this.

**Breaking change**: The name of the argument is now `image` instead of
`images` since it expects by default only 1 image. This is revertable
like the previous one.

**Breaking change**: The types is now simplified and flattened:

`pipe(inputs) == [{**object1}, {**object2}]`
instead of the previous
`pipe(inputs) == [[{**object1}, {**object1}], [{**object2}]]`
Where the different instances would be grouped by candidate labels
within lists.
IMHO this is not really desirable, since it would output empty lists and
is only adding superflous indirection compared to
`zero-shot-object-detection`.

It is relatively change free in terms of how the results, it does change
computation however since now the batching is handled by the pipeline
itself. It **did** change the results for the small models so there
seems to be a real difference in how the models handle this.

* Fixing the doctests.

* Behind is_torch_available.
This commit is contained in:
Nicolas Patry
2022-11-18 15:57:28 +01:00
committed by GitHub
parent 84c9cc6d15
commit 8e777b3ba4
2 changed files with 133 additions and 255 deletions
Download Patch File Download Diff File Expand all files Collapse all files

148
tests/pipelines/test_pipelines_zero_shot_object_detection.py
View File

@@ -43,28 +43,28 @@ class ZeroShotObjectDetectionPipelineTests(unittest.TestCase, metaclass=Pipeline
examples = [
{
"images": "./tests/fixtures/tests_samples/COCO/000000039769.png",
"text_queries": ["cat", "remote", "couch"],
"image": "./tests/fixtures/tests_samples/COCO/000000039769.png",
"candidate_labels": ["cat", "remote", "couch"],
}
]
return object_detector, examples
def run_pipeline_test(self, object_detector, examples):
batch_outputs = object_detector(examples, threshold=0.0)
outputs = object_detector(examples[0], threshold=0.0)
self.assertEqual(len(examples), len(batch_outputs))
for outputs in batch_outputs:
for output_per_image in outputs:
self.assertGreater(len(output_per_image), 0)
for detected_object in output_per_image:
self.assertEqual(
detected_object,
{
"score": ANY(float),
"label": ANY(str),
"box": {"xmin": ANY(int), "ymin": ANY(int), "xmax": ANY(int), "ymax": ANY(int)},
},
)
n = len(outputs)
self.assertGreater(n, 0)
self.assertEqual(
outputs,
[
{
"score": ANY(float),
"label": ANY(str),
"box": {"xmin": ANY(int), "ymin": ANY(int), "xmax": ANY(int), "ymax": ANY(int)},
}
for i in range(n)
],
)
@require_tf
@unittest.skip("Zero Shot Object Detection not implemented in TF")
@@ -79,64 +79,32 @@ class ZeroShotObjectDetectionPipelineTests(unittest.TestCase, metaclass=Pipeline
outputs = object_detector(
"./tests/fixtures/tests_samples/COCO/000000039769.png",
text_queries=["cat", "remote", "couch"],
candidate_labels=["cat", "remote", "couch"],
threshold=0.64,
)
self.assertEqual(
nested_simplify(outputs, decimals=4),
[
[
{"score": 0.7235, "label": "cat", "box": {"xmin": 204, "ymin": 167, "xmax": 232, "ymax": 190}},
{"score": 0.6748, "label": "remote", "box": {"xmin": 571, "ymin": 83, "xmax": 598, "ymax": 103}},
{"score": 0.6456, "label": "remote", "box": {"xmin": 494, "ymin": 105, "xmax": 521, "ymax": 127}},
{"score": 0.642, "label": "remote", "box": {"xmin": 67, "ymin": 274, "xmax": 93, "ymax": 297}},
]
],
)
outputs = object_detector(
["./tests/fixtures/tests_samples/COCO/000000039769.png"],
text_queries=["cat", "remote", "couch"],
threshold=0.64,
)
self.assertEqual(
nested_simplify(outputs, decimals=4),
[
[
{"score": 0.7235, "label": "cat", "box": {"xmin": 204, "ymin": 167, "xmax": 232, "ymax": 190}},
{"score": 0.6748, "label": "remote", "box": {"xmin": 571, "ymin": 83, "xmax": 598, "ymax": 103}},
{"score": 0.6456, "label": "remote", "box": {"xmin": 494, "ymin": 105, "xmax": 521, "ymax": 127}},
{"score": 0.642, "label": "remote", "box": {"xmin": 67, "ymin": 274, "xmax": 93, "ymax": 297}},
]
],
)
outputs = object_detector(
"./tests/fixtures/tests_samples/COCO/000000039769.png",
text_queries=[["cat", "remote", "couch"]],
threshold=0.64,
)
self.assertEqual(
nested_simplify(outputs, decimals=4),
[
[
{"score": 0.7235, "label": "cat", "box": {"xmin": 204, "ymin": 167, "xmax": 232, "ymax": 190}},
{"score": 0.6748, "label": "remote", "box": {"xmin": 571, "ymin": 83, "xmax": 598, "ymax": 103}},
{"score": 0.6456, "label": "remote", "box": {"xmin": 494, "ymin": 105, "xmax": 521, "ymax": 127}},
{"score": 0.642, "label": "remote", "box": {"xmin": 67, "ymin": 274, "xmax": 93, "ymax": 297}},
]
{"score": 0.7235, "label": "cat", "box": {"xmin": 204, "ymin": 167, "xmax": 232, "ymax": 190}},
{"score": 0.7218, "label": "remote", "box": {"xmin": 204, "ymin": 167, "xmax": 232, "ymax": 190}},
{"score": 0.7184, "label": "couch", "box": {"xmin": 204, "ymin": 167, "xmax": 232, "ymax": 190}},
{"score": 0.6748, "label": "remote", "box": {"xmin": 571, "ymin": 83, "xmax": 598, "ymax": 103}},
{"score": 0.6656, "label": "cat", "box": {"xmin": 571, "ymin": 83, "xmax": 598, "ymax": 103}},
{"score": 0.6614, "label": "couch", "box": {"xmin": 571, "ymin": 83, "xmax": 598, "ymax": 103}},
{"score": 0.6456, "label": "remote", "box": {"xmin": 494, "ymin": 105, "xmax": 521, "ymax": 127}},
{"score": 0.642, "label": "remote", "box": {"xmin": 67, "ymin": 274, "xmax": 93, "ymax": 297}},
{"score": 0.6419, "label": "cat", "box": {"xmin": 494, "ymin": 105, "xmax": 521, "ymax": 127}},
],
)
outputs = object_detector(
[
"./tests/fixtures/tests_samples/COCO/000000039769.png",
"http://images.cocodataset.org/val2017/000000039769.jpg",
{
"image": "./tests/fixtures/tests_samples/COCO/000000039769.png",
"candidate_labels": ["cat", "remote", "couch"],
}
],
text_queries=[["cat", "remote", "couch"], ["cat", "remote", "couch"]],
threshold=0.64,
)
@@ -145,16 +113,15 @@ class ZeroShotObjectDetectionPipelineTests(unittest.TestCase, metaclass=Pipeline
[
[
{"score": 0.7235, "label": "cat", "box": {"xmin": 204, "ymin": 167, "xmax": 232, "ymax": 190}},
{"score": 0.7218, "label": "remote", "box": {"xmin": 204, "ymin": 167, "xmax": 232, "ymax": 190}},
{"score": 0.7184, "label": "couch", "box": {"xmin": 204, "ymin": 167, "xmax": 232, "ymax": 190}},
{"score": 0.6748, "label": "remote", "box": {"xmin": 571, "ymin": 83, "xmax": 598, "ymax": 103}},
{"score": 0.6656, "label": "cat", "box": {"xmin": 571, "ymin": 83, "xmax": 598, "ymax": 103}},
{"score": 0.6614, "label": "couch", "box": {"xmin": 571, "ymin": 83, "xmax": 598, "ymax": 103}},
{"score": 0.6456, "label": "remote", "box": {"xmin": 494, "ymin": 105, "xmax": 521, "ymax": 127}},
{"score": 0.642, "label": "remote", "box": {"xmin": 67, "ymin": 274, "xmax": 93, "ymax": 297}},
],
[
{"score": 0.7235, "label": "cat", "box": {"xmin": 204, "ymin": 167, "xmax": 232, "ymax": 190}},
{"score": 0.6748, "label": "remote", "box": {"xmin": 571, "ymin": 83, "xmax": 598, "ymax": 103}},
{"score": 0.6456, "label": "remote", "box": {"xmin": 494, "ymin": 105, "xmax": 521, "ymax": 127}},
{"score": 0.642, "label": "remote", "box": {"xmin": 67, "ymin": 274, "xmax": 93, "ymax": 297}},
],
{"score": 0.6419, "label": "cat", "box": {"xmin": 494, "ymin": 105, "xmax": 521, "ymax": 127}},
]
],
)
@@ -164,27 +131,30 @@ class ZeroShotObjectDetectionPipelineTests(unittest.TestCase, metaclass=Pipeline
object_detector = pipeline("zero-shot-object-detection")
outputs = object_detector(
"http://images.cocodataset.org/val2017/000000039769.jpg", text_queries=["cat", "remote", "couch"]
"http://images.cocodataset.org/val2017/000000039769.jpg", candidate_labels=["cat", "remote", "couch"]
)
self.assertEqual(
nested_simplify(outputs, decimals=4),
[
[
{"score": 0.2868, "label": "cat", "box": {"xmin": 324, "ymin": 20, "xmax": 640, "ymax": 373}},
{"score": 0.277, "label": "remote", "box": {"xmin": 40, "ymin": 72, "xmax": 177, "ymax": 115}},
{"score": 0.2537, "label": "cat", "box": {"xmin": 1, "ymin": 55, "xmax": 315, "ymax": 472}},
{"score": 0.1474, "label": "remote", "box": {"xmin": 335, "ymin": 74, "xmax": 371, "ymax": 187}},
{"score": 0.1208, "label": "couch", "box": {"xmin": 4, "ymin": 0, "xmax": 642, "ymax": 476}},
]
{"score": 0.2868, "label": "cat", "box": {"xmin": 324, "ymin": 20, "xmax": 640, "ymax": 373}},
{"score": 0.277, "label": "remote", "box": {"xmin": 40, "ymin": 72, "xmax": 177, "ymax": 115}},
{"score": 0.2537, "label": "cat", "box": {"xmin": 1, "ymin": 55, "xmax": 315, "ymax": 472}},
{"score": 0.1474, "label": "remote", "box": {"xmin": 335, "ymin": 74, "xmax": 371, "ymax": 187}},
{"score": 0.1208, "label": "couch", "box": {"xmin": 4, "ymin": 0, "xmax": 642, "ymax": 476}},
],
)
outputs = object_detector(
[
"http://images.cocodataset.org/val2017/000000039769.jpg",
"http://images.cocodataset.org/val2017/000000039769.jpg",
{
"image": "http://images.cocodataset.org/val2017/000000039769.jpg",
"candidate_labels": ["cat", "remote", "couch"],
},
{
"image": "http://images.cocodataset.org/val2017/000000039769.jpg",
"candidate_labels": ["cat", "remote", "couch"],
},
],
text_queries=[["cat", "remote", "couch"], ["cat", "remote", "couch"]],
)
self.assertEqual(
nested_simplify(outputs, decimals=4),
@@ -219,17 +189,15 @@ class ZeroShotObjectDetectionPipelineTests(unittest.TestCase, metaclass=Pipeline
outputs = object_detector(
"http://images.cocodataset.org/val2017/000000039769.jpg",
text_queries=["cat", "remote", "couch"],
candidate_labels=["cat", "remote", "couch"],
threshold=threshold,
)
self.assertEqual(
nested_simplify(outputs, decimals=4),
[
[
{"score": 0.2868, "label": "cat", "box": {"xmin": 324, "ymin": 20, "xmax": 640, "ymax": 373}},
{"score": 0.277, "label": "remote", "box": {"xmin": 40, "ymin": 72, "xmax": 177, "ymax": 115}},
{"score": 0.2537, "label": "cat", "box": {"xmin": 1, "ymin": 55, "xmax": 315, "ymax": 472}},
]
{"score": 0.2868, "label": "cat", "box": {"xmin": 324, "ymin": 20, "xmax": 640, "ymax": 373}},
{"score": 0.277, "label": "remote", "box": {"xmin": 40, "ymin": 72, "xmax": 177, "ymax": 115}},
{"score": 0.2537, "label": "cat", "box": {"xmin": 1, "ymin": 55, "xmax": 315, "ymax": 472}},
],
)
@@ -241,15 +209,13 @@ class ZeroShotObjectDetectionPipelineTests(unittest.TestCase, metaclass=Pipeline
outputs = object_detector(
"http://images.cocodataset.org/val2017/000000039769.jpg",
text_queries=["cat", "remote", "couch"],
candidate_labels=["cat", "remote", "couch"],
top_k=top_k,
)
self.assertEqual(
nested_simplify(outputs, decimals=4),
[
[
{"score": 0.2868, "label": "cat", "box": {"xmin": 324, "ymin": 20, "xmax": 640, "ymax": 373}},
{"score": 0.277, "label": "remote", "box": {"xmin": 40, "ymin": 72, "xmax": 177, "ymax": 115}},
]
{"score": 0.2868, "label": "cat", "box": {"xmin": 324, "ymin": 20, "xmax": 640, "ymax": 373}},
{"score": 0.277, "label": "remote", "box": {"xmin": 40, "ymin": 72, "xmax": 177, "ymax": 115}},
],
)