Detectron2

Detectron2

Detectron2 is Meta AI Research's next-generation platform for object detection and segmentation. It implements state-of-the-art algorithms including Faster R-CNN, Mask R-CNN, RetinaNet, DETR, PointRend, and Panoptic-FPN in a modular, extensible framework backed by the Detectron2 config system.

Key Features

• Pre-trained models for detection, instance/panoptic/semantic segmentation, and pose estimation
• Highly modular: swap backbone, neck, head, and loss independently
• Mixed-precision training and distributed training across multiple GPUs/nodes
• Lazy config system for cleaner experiment management
• Tight integration with COCO, LVIS, Cityscapes, and custom datasets

Quick Start

pip install 'git+https://github.com/facebookresearch/detectron2.git'

from detectron2 import model_zoo
from detectron2.engine import DefaultPredictor
from detectron2.config import get_cfg
import cv2

cfg = get_cfg()
cfg.merge_from_file(model_zoo.get_config_file(
    "COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_3x.yaml"))
cfg.MODEL.WEIGHTS = model_zoo.get_checkpoint_url(
    "COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_3x.yaml")
cfg.MODEL.ROI_HEADS.SCORE_THRESH_TEST = 0.5
predictor = DefaultPredictor(cfg)

img = cv2.imread("input.jpg")
outputs = predictor(img)
print(outputs["instances"].pred_classes)

npx ai-supply add detectron2-object-detection

Curated mirror of the open-source Detectron2 (Apache-2.0). Get it from the source.