Spaces:
Runtime error
Runtime error
File size: 2,253 Bytes
c3a1897 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 |
import argparse
import multiprocessing as mp
import os
import time
import cv2
import tqdm
import sys
from detectron2.config import get_cfg
from detectron2.data.detection_utils import read_image
from detectron2.utils.logger import setup_logger
sys.path.insert(0, 'models/grit_src/third_party/CenterNet2/projects/CenterNet2/')
from centernet.config import add_centernet_config
from models.grit_src.grit.config import add_grit_config
from models.grit_src.grit.predictor import VisualizationDemo
import json
# constants
WINDOW_NAME = "GRiT"
def dense_pred_to_caption(predictions):
boxes = predictions["instances"].pred_boxes if predictions["instances"].has("pred_boxes") else None
object_description = predictions["instances"].pred_object_descriptions.data
new_caption = ""
for i in range(len(object_description)):
new_caption += (object_description[i] + ": " + str([int(a) for a in boxes[i].tensor.cpu().detach().numpy()[0]])) + "; "
return new_caption
def setup_cfg(args):
cfg = get_cfg()
if args["cpu"]:
cfg.MODEL.DEVICE="cpu"
add_centernet_config(cfg)
add_grit_config(cfg)
cfg.merge_from_file(args["config_file"])
cfg.merge_from_list(args["opts"])
# Set score_threshold for builtin models
cfg.MODEL.ROI_HEADS.SCORE_THRESH_TEST = args["confidence_threshold"]
cfg.MODEL.PANOPTIC_FPN.COMBINE.INSTANCES_CONFIDENCE_THRESH = args["confidence_threshold"]
if args["test_task"]:
cfg.MODEL.TEST_TASK = args["test_task"]
cfg.MODEL.BEAM_SIZE = 1
cfg.MODEL.ROI_HEADS.SOFT_NMS_ENABLED = False
cfg.USE_ACT_CHECKPOINT = False
cfg.freeze()
return cfg
def get_parser():
arg_dict = {'config_file': "models/grit_src/configs/GRiT_B_DenseCap_ObjectDet.yaml", 'cpu': False, 'confidence_threshold': 0.5, 'test_task': 'DenseCap', 'opts': ["MODEL.WEIGHTS", "pretrained_models/grit_b_densecap_objectdet.pth"]}
return arg_dict
def image_caption_api(image_src):
args2 = get_parser()
cfg = setup_cfg(args2)
demo = VisualizationDemo(cfg)
if image_src:
img = read_image(image_src, format="BGR")
predictions, visualized_output = demo.run_on_image(img)
new_caption = dense_pred_to_caption(predictions)
return new_caption |