Spaces:
Runtime error
Runtime error
File size: 8,579 Bytes
e98e70e 8d6560f e98e70e 5e262d5 e98e70e 8d6560f e98e70e c983071 e98e70e 8d6560f 58e86c5 8d6560f e98e70e 8d6560f e98e70e 8d6560f e98e70e 8d6560f ba61a77 e98e70e 5e262d5 e98e70e 5e262d5 e98e70e 8d6560f e98e70e d4c1aca e98e70e 8d6560f e98e70e d4c1aca e98e70e d4c1aca e98e70e 8d6560f e98e70e 5e262d5 e98e70e 5e262d5 e98e70e 2d2a6a7 |
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 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 |
import torch
import gradio as gr
import cv2
import numpy as np
import random
import numpy as np
from models.experimental import attempt_load
from utils.general import check_img_size, non_max_suppression, \
scale_coords
from utils.plots import plot_one_box
from utils.torch_utils import time_synchronized
import time
from ultralytics import YOLO
def letterbox(im, new_shape=(640, 640), color=(114, 114, 114), auto=True, scaleup=True, stride=32):
# Resize and pad image while meeting stride-multiple constraints
shape = im.shape[:2] # current shape [height, width]
if isinstance(new_shape, int):
new_shape = (new_shape, new_shape)
# Scale ratio (new / old)
r = min(new_shape[0] / shape[0], new_shape[1] / shape[1])
if not scaleup: # only scale down, do not scale up (for better val mAP)
r = min(r, 1.0)
# Compute padding
new_unpad = int(round(shape[1] * r)), int(round(shape[0] * r))
dw, dh = new_shape[1] - new_unpad[0], new_shape[0] - new_unpad[1] # wh padding
if auto: # minimum rectangle
dw, dh = np.mod(dw, stride), np.mod(dh, stride) # wh padding
dw /= 2 # divide padding into 2 sides
dh /= 2
if shape[::-1] != new_unpad: # resize
im = cv2.resize(im, new_unpad, interpolation=cv2.INTER_LINEAR)
top, bottom = int(round(dh - 0.1)), int(round(dh + 0.1))
left, right = int(round(dw - 0.1)), int(round(dw + 0.1))
im = cv2.copyMakeBorder(im, top, bottom, left, right, cv2.BORDER_CONSTANT, value=color) # add border
return im, r, (dw, dh)
names = ["animal",
"autorickshaw",
"bicycle",
"bus",
"car",
"motorcycle",
"person",
"rider",
"traffic light",
"traffic sign",
"truck"
]
#colors = [[random.randint(0, 255) for _ in range(3)] for _ in names]
colors = {
"animal": [246,198, 145],
"autorickshaw": [255,204, 54],
"bicycle": [119,11, 32],
"bus": [ 0,60,100],
"car": [ 0,0,142],
"motorcycle": [ 0,0,230],
"person": [220,20, 60],
"rider": [255,0, 0],
"traffic light": [250,170, 30],
"traffic sign": [220,220, 0],
"truck": [ 0,0, 70]
}
def detectv7(img,model,device,iou_threshold=0.45,confidence_threshold=0.25):
imgsz = 640
img = np.array(img)
stride = int(model.stride.max()) # model stride
imgsz = check_img_size(imgsz, s=stride) # check img_size
# Get names and colors
names = model.module.names if hasattr(model, 'module') else model.names
# Run inference
imgs = img.copy() # for NMS
image, ratio, dwdh = letterbox(img, auto=False)
image = image.transpose((2, 0, 1))
img = torch.from_numpy(image).to(device)
img = img.float() # uint8 to fp16/32
img /= 255.0 # 0 - 255 to 0.0 - 1.0
if img.ndimension() == 3:
img = img.unsqueeze(0)
# Inference
t1 = time_synchronized()
start = time.time()
with torch.no_grad(): # Calculating gradients would cause a GPU memory leak
pred = model(img,augment=True)[0]
fps_inference = 1/(time.time()-start)
t2 = time_synchronized()
# Apply NMS
pred = non_max_suppression(pred, confidence_threshold, iou_threshold, classes=None, agnostic=True)
t3 = time_synchronized()
for i, det in enumerate(pred): # detections per image
if len(det):
# Rescale boxes from img_size to im0 size
det[:, :4] = scale_coords(img.shape[2:], det[:, :4], imgs.shape).round()
# Write results
for *xyxy, conf, cls in reversed(det):
label = f'{names[int(cls)]} {conf:.2f}'
plot_one_box(xyxy, imgs, label=label, color=colors[names[int(cls)]], line_thickness=2)
return imgs,fps_inference
def detectv8(img,model,device,iou_threshold=0.45,confidence_threshold=0.25):
img = np.array(img)
# Inference
t1 = time_synchronized()
start = time.time()
results= model.predict(img,conf=confidence_threshold, iou=iou_threshold)
fps_inference = 1/(time.time()-start)
boxes=results[0].boxes.numpy()
for bbox in boxes:
#print(f'{colors[names[int(bbox.cls[0])]]}')
label = f'{names[int(bbox.cls[0])]} {bbox.conf[0]:.2f}'
plot_one_box(bbox.xyxy[0],img,colors[names[int(bbox.cls[0])]],label, line_thickness=1)
return img,fps_inference
def inference(img,model_link,iou_threshold,confidence_threshold):
print(model_link)
device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')
# Load model
model_path = 'weights/'+str(model_link)+'.pt'
if model_link== 'yolov8m':
model = YOLO(model_path)
return detectv8(img,model,device,iou_threshold,confidence_threshold)
else:
model = attempt_load(model_path, map_location=device)
return detectv7(img,model,device,iou_threshold,confidence_threshold)
def inference2(video,model_link,iou_threshold,confidence_threshold):
print(model_link)
device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')
# Load model
model_path = 'weights/'+str(model_link)+'.pt'
if model_link== 'yolov8m':
model = YOLO(model_path)
else:
model = attempt_load(model_path, map_location=device)
frames = cv2.VideoCapture(video)
fps = frames.get(cv2.CAP_PROP_FPS)
image_size = (int(frames.get(cv2.CAP_PROP_FRAME_WIDTH)),int(frames.get(cv2.CAP_PROP_FRAME_HEIGHT)))
finalVideo = cv2.VideoWriter('output.mp4',cv2.VideoWriter_fourcc(*'VP90'), fps, image_size)
fps_video = []
while frames.isOpened():
ret,frame = frames.read()
if not ret:
break
if model_link== 'yolov8m':
frame,fps = detectv8(frame,model,device,iou_threshold,confidence_threshold)
else:
frame,fps = detectv7(frame,model,device,iou_threshold,confidence_threshold)
fps_video.append(fps)
finalVideo.write(frame)
frames.release()
finalVideo.release()
return 'output.mp4',np.mean(fps_video)
examples_images = ['data/images/1.jpg',
'data/images/2.jpg',
'data/images/bus.jpg',
'data/images/3.jpg']
examples_videos = ['data/video/1.mp4','data/video/2.mp4']
models = ['yolov8m','yolov7','yolov7t']
def update_markdown(value):
return f"## {value} Inference on Image"
with gr.Blocks() as demo:
gr.Markdown("## IDD Inference on Yolo V7 and V8 ")
with gr.Tab("Image"):
im_md= gr.Markdown("## Yolo V7 and V8 Inference on Image")
with gr.Row():
image_input = gr.Image(type='pil', label="Input Image", source="upload")
image_output = gr.Image(type='pil', label="Output Image", source="upload")
fps_image = gr.Number(0,label='FPS')
image_drop = gr.Dropdown(choices=models,value=models[0])
image_iou_threshold = gr.Slider(label="IOU Threshold",interactive=True, minimum=0.0, maximum=1.0, value=0.45)
image_conf_threshold = gr.Slider(label="Confidence Threshold",interactive=True, minimum=0.0, maximum=1.0, value=0.25)
gr.Examples(examples=examples_images,inputs=image_input,outputs=image_output)
text_button = gr.Button("Detect")
image_drop.change(update_markdown,image_drop,im_md)
with gr.Tab("Video"):
gr.Markdown("## Yolo V7 and V8 Inference on Video")
with gr.Row():
video_input = gr.Video(type='pil', label="Input Video", source="upload")
video_output = gr.Video(type="pil", label="Output Video",format="mp4")
fps_video = gr.Number(0,label='FPS')
video_drop = gr.Dropdown(choices=models,value=models[0])
video_iou_threshold = gr.Slider(label="IOU Threshold",interactive=True, minimum=0.0, maximum=1.0, value=0.45)
video_conf_threshold = gr.Slider(label="Confidence Threshold",interactive=True, minimum=0.0, maximum=1.0, value=0.25)
gr.Examples(examples=examples_videos,inputs=video_input,outputs=video_output)
video_button = gr.Button("Detect")
# with gr.Tab("Webcam Video"):
# gr.Markdown("## YOLOv7 Inference on Webcam Video")
# gr.Markdown("Coming Soon")
text_button.click(inference, inputs=[image_input,image_drop,
image_iou_threshold,image_conf_threshold],
outputs=[image_output,fps_image])
video_button.click(inference2, inputs=[video_input,video_drop,
video_iou_threshold,video_conf_threshold],
outputs=[video_output,fps_video])
demo.launch(debug=True,enable_queue=True) |