app.py

import os

os.system("pip install gradio==2.4.6")
import sys
import gradio as gr

os.system(
    "pip install detectron2 -f https://dl.fbaipublicfiles.com/detectron2/wheels/cu102/torch1.9/index.html"
)

# clone and install Detic
os.system(
    "git clone https://github.com/facebookresearch/Detic.git --recurse-submodules"
)
os.chdir("Detic")

# Install detectron2
import torch

# Some basic setup:
# Setup detectron2 logger
import detectron2
from detectron2.utils.logger import setup_logger

setup_logger()

# import some common libraries
import sys
import numpy as np
import os, json, cv2, random

# import some common detectron2 utilities
from detectron2 import model_zoo
from detectron2.engine import DefaultPredictor
from detectron2.config import get_cfg
from detectron2.utils.visualizer import Visualizer
from detectron2.data import MetadataCatalog, DatasetCatalog

# Detic libraries
sys.path.insert(0, "third_party/CenterNet2/projects/CenterNet2/")
sys.path.insert(0, "third_party/CenterNet2/")
from centernet.config import add_centernet_config
from detic.config import add_detic_config
from detic.modeling.utils import reset_cls_test

from PIL import Image

# Build the detector and download our pretrained weights
cfg = get_cfg()
add_centernet_config(cfg)
add_detic_config(cfg)
cfg.MODEL.DEVICE = "cpu"
cfg.merge_from_file("configs/Detic_LCOCOI21k_CLIP_SwinB_896b32_4x_ft4x_max-size.yaml")
cfg.MODEL.WEIGHTS = "https://dl.fbaipublicfiles.com/detic/Detic_LCOCOI21k_CLIP_SwinB_896b32_4x_ft4x_max-size.pth"
cfg.MODEL.ROI_HEADS.SCORE_THRESH_TEST = 0.5  # set threshold for this model
cfg.MODEL.ROI_BOX_HEAD.ZEROSHOT_WEIGHT_PATH = "rand"
cfg.MODEL.ROI_HEADS.ONE_CLASS_PER_PROPOSAL = (
    True  # For better visualization purpose. Set to False for all classes.
)
predictor = DefaultPredictor(cfg)

# Setup the model's vocabulary using build-in datasets

BUILDIN_CLASSIFIER = {
    "lvis": "datasets/metadata/lvis_v1_clip_a+cname.npy",
    "objects365": "datasets/metadata/o365_clip_a+cnamefix.npy",
    "openimages": "datasets/metadata/oid_clip_a+cname.npy",
    "coco": "datasets/metadata/coco_clip_a+cname.npy",
}

BUILDIN_METADATA_PATH = {
    "lvis": "lvis_v1_val",
    "objects365": "objects365_v2_val",
    "openimages": "oid_val_expanded",
    "coco": "coco_2017_val",
}

vocabulary = "lvis"  # change to 'lvis', 'objects365', 'openimages', or 'coco'
metadata = MetadataCatalog.get(BUILDIN_METADATA_PATH[vocabulary])
classifier = BUILDIN_CLASSIFIER[vocabulary]
num_classes = len(metadata.thing_classes)
reset_cls_test(predictor.model, classifier, num_classes)

os.system("wget https://web.eecs.umich.edu/~fouhey/fun/desk/desk.jpg")


def inference(img):

    im = cv2.imread(img)

    outputs = predictor(im)
    v = Visualizer(im[:, :, ::-1], metadata)
    out = v.draw_instance_predictions(outputs["instances"].to("cpu"))

    detected_objects = []
    box_locations = outputs["instances"].pred_boxes
    box_loc_screen = box_locations.tensor.cpu().numpy()

    for i, box_coord in enumerate(box_loc_screen):
        x0, y0, x1, y1 = box_coord
        width = x1 - x0
        height = y1 - y0
        predicted_label = metadata.thing_classes[outputs["instances"].pred_classes[i]]
        detected_objects.append(
            {
                "prediction": predicted_label,
                "x": int(x0),
                "y": int(y0),
                "w": int(width),
                "h": int(height),
            }
        )

    return Image.fromarray(np.uint8(out.get_image())).convert("RGB"), detected_objects


title = "Detic"

description = "Gradio demo for Detic: Detecting Twenty-thousand Classes using Image-level Supervision. To use it, simply upload your image, or click one of the examples to load them. Read more at the links below."

article = "<p style='text-align: center'><a href='https://arxiv.org/abs/2201.02605' target='_blank'>Detecting Twenty-thousand Classes using Image-level Supervision</a> | <a href='https://github.com/facebookresearch/Detic' target='_blank'>Github Repo</a></p>"

examples = [["desk.jpg"]]
gr.Interface(
    inference,
    inputs=gr.inputs.Image(type="filepath"),
    outputs=[
        gr.outputs.Image(label="Visualization", type="pil"),
        gr.outputs.JSON(label="Detected Objects"),
    ],
    enable_queue=True,
    title=title,
    description=description,
    article=article,
    examples=examples,
).launch()