autodistill / autodistill-codet

CoDet base model for use with Autodistill.

Autodistill CoDet Module

This repository contains the code supporting the CoDet base model for use with Autodistill.

CoDet is an open vocabulary zero-shot object detection model. The model was described in the "CoDet: Co-Occurrence Guided Region-Word Alignment for Open-Vocabulary Object Detection" published by Chuofan Ma, Yi Jiang, Xin Wen, Zehuan Yuan, Xiaojuan Qi. The paper was submitted to NeurIPS2023.

Read the full Autodistill documentation.

Read the CoDet Autodistill documentation.

Installation

To use CoDet with autodistill, you need to install the following dependency:

pip3 install autodistill-codet

Quickstart

When you first run the model, it will download CoDet and its dependencies, as well as the required model configuration and weights. The output during the download process will be verbose. If you stop the download process before it has finished, run rm -rf ~/.cache/autodistill/CoDet before running the model again. This ensures that you don't work from a part-installed CoDet setup.

When the predict() function runs, the output will also be verbose. You can ignore the output printed to the console that appears when you call predict().

You can only predict classes in the LVIS vocabulary. You can see a list of supported classes in the class_names.json file in the autodistill-codet GitHub repository.

Use the code snippet below to get started:

from autodistill_codet import CoDet
from autodistill.detection import CaptionOntology
from autodistill.utils import plot
import cv2

# define an ontology to map class names to our CoDet prompt
# the ontology dictionary has the format {caption: class}
# where caption is the prompt sent to the base model, and class is the label that will
# be saved for that caption in the generated annotations
# then, load the model
base_model = CoDet(
    ontology=CaptionOntology(
        {
            "person": "person"
        }
    )
)

# run inference on an image and display the results
# class_names is a list of all classes supported by the model
# class_names can be used to turn the class_id values from the model into human-readable class names
# class names is defined in self.class_names
predictions = base_model.predict("./context_images/1.jpeg")
image = cv2.imread("./context_images/1.jpeg")

plot(
  image=image,
  detections=predictions,
  classes=base_model.class_names
)

# run inference on a folder of images and save the results
base_model.label("./context_images", extension=".jpeg")

License

This project is licensed under an Apache 2.0 license, except where files explicitly note a license.

🏆 Contributing

We love your input! Please see the core Autodistill contributing guide to get started. Thank you 🙏 to all our contributors!

autodistill-codet's Issues

ModuleNotFoundError: No module named 'detectron2'

Hello,

I am trying to install the CO-Det implementation with Autodistill. I had installed autodistill with the given command-line in the the documentation:

pip install autodistill autodistill-grounded-sam autodistill-yolov8

I am running the following demo script:

from autodistill_codet import CoDet
from autodistill.detection import CaptionOntology
from autodistill.utils import plot
import cv2

base_model = CoDet(ontology=CaptionOntology({"person": "person"}))

image_path = "../my-img.jpg"
predictions = base_model.predict(image_path)
image = cv2.imread(image_path)

plot(image=image, detections=predictions, classes=base_model.class_names)

I am getting the following error:

Traceback (most recent call last):
  File "/home/xx/Desktop/auto-labeling-models-evaluation/codet/preview.py", line 1, in <module>
    from autodistill_codet import CoDet
  File "/home/xx/miniconda3/envs/autodistill/lib/python3.11/site-packages/autodistill_codet/__init__.py", line 1, in <module>
    from autodistill_codet.codet_model import CoDet
  File "/home/xx/miniconda3/envs/autodistill/lib/python3.11/site-packages/autodistill_codet/codet_model.py", line 86, in <module>
    import detectron2.utils.comm as comm
ModuleNotFoundError: No module named 'detectron2'

I already tried to install detectron2 separately but the CO-Det install overwrite it anyaway.

Thank you for your help,
Rivoks