Emergent Object Detection/Classification Model.v1 #56
Conversation
noahbodine
added
vision
for more information, see https://pre-commit.ci
There was a problem hiding this comment.
Need a few changes to make sure the code is up to repo standards. You can check out the comments below for specifics. In addition to those comments, a few more changes below:
- Might want to choose a better name for
best.pt. Something likeemergent_model.ptor weights or something - Make sure that your code is passing pre-commit checks for mypy and pylint. You can check how to install/use these in the
ReadMe.md
|
Good first issue for someone to take up the fixes for this PR |
fallscameron01
modified the milestones:
Emergent Object Classification,
Emergent Object Detection
|
I worked out how to use the model and the output and added some code demonstrating its use. Docstrings and proper typehinting still need to be added, but the functionality should be pretty much all there at this point. |
| # You may have to install: | ||
| # pandas | ||
| # torchvision | ||
| # tqdm | ||
| # seaborn |
There was a problem hiding this comment.
Dependencies should all be taken care of by #58 if you run in a poetry shell. These comments can be removed.
|
|
||
| MODEL_PATH = "vision/emergent_object/best.pt" |
There was a problem hiding this comment.
Since this constant is only used in emergent_object_model(), constant MODEL_PATH should be moved to that function.
| # Function to do detection / classification | ||
| def emergent_object_detection(image, model): | ||
| # Convert to RGB | ||
| image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB) | ||
|
|
||
| # Run the model on the image. Both a file path and a numpy image work, but | ||
| # we want to use a numpy image | ||
| results = model(image) | ||
|
|
||
| # Retrieve the output from the model | ||
| output = results.pandas().xyxy[0] | ||
|
|
||
| return output |
There was a problem hiding this comment.
Function needs:
- docstring
- parameter types
- return type
| # Convert to RGB | ||
| image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB) |
There was a problem hiding this comment.
Type hint for image should be Image type alias from vision > common > constants.py
| results = model(image) | ||
|
|
||
| # Retrieve the output from the model | ||
| output = results.pandas().xyxy[0] |
There was a problem hiding this comment.
type hints for results and output
| # Convert the Pandas Dataframe to a dictionary - this will be necessary and | ||
| # should eventually be done in `emergent_object_detection()` | ||
| output_dict = output.to_dict("index") |
There was a problem hiding this comment.
type hint for output_dict
| # Draw the bounding boxes to the original image | ||
| for row in output_dict.values(): |
There was a problem hiding this comment.
Type hint for row. Loop variables are type hinted like this:
row: Type
for row ...
| # Get the output ranges | ||
| top_left = (int(row["xmin"]), int(row["ymin"])) | ||
| bottom_right = (int(row["xmax"]), int(row["ymax"])) |
There was a problem hiding this comment.
type hints for these two variables
| # Display the image | ||
| cv2.imshow("", image) |
There was a problem hiding this comment.
Might want something like "Emergent Object" instead of an empty string for the title.
| # Function to do detection / classification | ||
| def emergent_object_detection(image, model): | ||
| # Convert to RGB | ||
| image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB) | ||
|
|
||
| # Run the model on the image. Both a file path and a numpy image work, but | ||
| # we want to use a numpy image | ||
| results = model(image) | ||
|
|
||
| # Retrieve the output from the model | ||
| output = results.pandas().xyxy[0] | ||
|
|
||
| return output |
There was a problem hiding this comment.
For the docstring for this function, make sure to describe what kind of output one should expect from it.
|
Opened new PR #96 because something was wrong with this branch that caused a split. |
Closes #12
Model to detect / classify emergent objects & pretrained weights for the model.