Spaces:
Runtime error
Runtime error
| import csv | |
| import gradio as gr | |
| import glob | |
| import pprint as pp | |
| from sys import excepthook | |
| from re import T | |
| from urllib.parse import parse_qs, urlparse | |
| import clip | |
| import numpy as np | |
| import requests | |
| import torch | |
| import io | |
| from IPython.display import Image, display | |
| from PIL import Image, ImageFont | |
| import os | |
| import cv2 | |
| import torch | |
| import glob | |
| # Model | |
| def predict(img,text): | |
| import tempfile | |
| model = torch.hub.load('ultralytics/yolov5', 'yolov5s') | |
| results = model(img) | |
| dirpath = tempfile.mkdtemp() | |
| results.crop(save_dir=dirpath) | |
| path= dirpath+'/crops/**/*.jpg' | |
| txtfiles = [] | |
| for file in glob.glob(path): | |
| txtfiles.append(file) | |
| from PIL import Image | |
| l = [] | |
| #keyList = list(range(len(txtfiles))) | |
| for filename in glob.glob(path): | |
| foo = Image.open(filename).convert('RGB') | |
| #resized_image = foo.resize((250,250)) | |
| l.append(foo) | |
| device = "cuda" if torch.cuda.is_available() else "cpu" | |
| model, preprocess = clip.load("ViT-B/32", device=device) | |
| images = torch.stack([preprocess(im) for im in l]).to(device) | |
| with torch.no_grad(): | |
| image_features = model.encode_image(images) | |
| image_features /= image_features.norm(dim=-1, keepdim=True) | |
| image_features.cpu().numpy() | |
| image_mean = torch.tensor([0.48145466, 0.4578275, 0.40821073]) | |
| image_std = torch.tensor([0.26862954, 0.26130258, 0.27577711]) | |
| images = [preprocess(im) for im in l] | |
| image_input = torch.tensor(np.stack(images)) | |
| image_input -= image_mean[:, None, None] | |
| image_input /= image_std[:, None, None] | |
| with torch.no_grad(): | |
| image_features = model.encode_image(image_input).float() | |
| image_features /= image_features.norm(dim=-1, keepdim=True) | |
| def get_top_N_semantic_similarity(similarity_list,N): | |
| results = zip(range(len(similarity_list)), similarity_list) | |
| results = sorted(results, key=lambda x: x[1],reverse= True) | |
| top_N_images = [] | |
| scores=[] | |
| for index,score in results[:N]: | |
| scores.append(score) | |
| top_N_images.append(l[index]) | |
| return scores,top_N_images | |
| #search_query = text | |
| with torch.no_grad(): | |
| # Encode and normalize the description using CLIP | |
| text_encoded = model.encode_text(clip.tokenize(text).to(device)) | |
| text_encoded /= text_encoded.norm(dim=-1, keepdim=True) | |
| similarity = text_encoded.cpu().numpy() @ image_features.cpu().numpy().T | |
| similarity = similarity[0] | |
| scores,imgs= get_top_N_semantic_similarity(similarity,N=1) | |
| #print ("scores ",scores) | |
| #ipyplot.plot_images(imgs,img_width=350) | |
| return imgs[0] | |
| #text = gr.inputs.Textbox(lines=1, label="Text query", placeholder="Introduce the search text...",) | |
| #img = gr.inputs.Image() | |
| #img = "image" | |
| gr.Interface(predict, ["image", gr.inputs.Textbox(lines=1, label="Text query", placeholder="Type here...",)], outputs="image", title="Crop-CLIP", description ="Search subjects/objects in an image using simple text description and get cropped results.This is done by combining Object detection Yolov5 and OpenAI's CLIP model.").launch(); | |