Update app.py
Browse files
app.py
CHANGED
@@ -1,15 +1,8 @@
|
|
1 |
-
|
2 |
import gradio as gr
|
3 |
-
import
|
4 |
-
import torch
|
5 |
-
import torchvision
|
6 |
import numpy as np
|
7 |
-
import
|
8 |
-
import evaluate
|
9 |
-
from transformers import TrainingArguments, Trainer
|
10 |
-
from transformers import VideoMAEImageProcessor, VideoMAEForVideoClassification
|
11 |
-
from torchvision.transforms import Compose
|
12 |
-
from pytorchvideo.data.labeled_video_dataset import LabeledVideoDataset
|
13 |
from pytorchvideo.transforms import (
|
14 |
ApplyTransformToKey,
|
15 |
Normalize,
|
@@ -18,157 +11,130 @@ from pytorchvideo.transforms import (
|
|
18 |
ShortSideScale,
|
19 |
UniformTemporalSubsample,
|
20 |
)
|
21 |
-
|
22 |
from torchvision.transforms import (
|
23 |
Compose,
|
24 |
Lambda,
|
|
|
|
|
25 |
Resize,
|
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 |
-
height = width = image_processor.size["shortest_edge"]
|
119 |
-
else:
|
120 |
-
height = image_processor.size["height"]
|
121 |
-
width = image_processor.size["width"]
|
122 |
-
|
123 |
-
resize_to = (height, width)
|
124 |
-
num_frames_to_sample = model_config.num_frames
|
125 |
-
|
126 |
-
transform = Compose(
|
127 |
-
[
|
128 |
-
UniformTemporalSubsample(num_frames_to_sample),
|
129 |
-
Lambda(lambda x: x / 255.0),
|
130 |
-
Normalize(mean, std),
|
131 |
-
Resize(resize_to),
|
132 |
-
]
|
133 |
-
)
|
134 |
-
|
135 |
-
video_tensor = transform(video)
|
136 |
-
return video_tensor
|
137 |
-
|
138 |
-
def run_inference(model, image_processor, video):
|
139 |
-
"""Utility to run inference given a model and test video."""
|
140 |
-
preprocessed_video = preprocess_video(video, image_processor, model.config)
|
141 |
-
inputs = {"pixel_values": preprocessed_video.unsqueeze(0)}
|
142 |
-
|
143 |
-
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
144 |
-
inputs = {k: v.to(device) for k, v in inputs.items()}
|
145 |
-
model = model.to(device)
|
146 |
-
|
147 |
with torch.no_grad():
|
148 |
-
outputs =
|
149 |
logits = outputs.logits
|
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 |
-
|
|
|
|
1 |
+
import cv2
|
2 |
import gradio as gr
|
3 |
+
import imutils
|
|
|
|
|
4 |
import numpy as np
|
5 |
+
import torch
|
|
|
|
|
|
|
|
|
|
|
6 |
from pytorchvideo.transforms import (
|
7 |
ApplyTransformToKey,
|
8 |
Normalize,
|
|
|
11 |
ShortSideScale,
|
12 |
UniformTemporalSubsample,
|
13 |
)
|
|
|
14 |
from torchvision.transforms import (
|
15 |
Compose,
|
16 |
Lambda,
|
17 |
+
RandomCrop,
|
18 |
+
RandomHorizontalFlip,
|
19 |
Resize,
|
20 |
)
|
21 |
+
from transformers import VideoMAEFeatureExtractor, VideoMAEForVideoClassification
|
22 |
+
|
23 |
+
MODEL_CKPT = "latif98/videomae-base-finetuned-isl-numbers_aug"
|
24 |
+
DEVICE = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
25 |
+
|
26 |
+
MODEL = VideoMAEForVideoClassification.from_pretrained(MODEL_CKPT).to(DEVICE)
|
27 |
+
PROCESSOR = VideoMAEFeatureExtractor.from_pretrained(MODEL_CKPT)
|
28 |
+
|
29 |
+
RESIZE_TO = PROCESSOR.size["shortest_edge"]
|
30 |
+
NUM_FRAMES_TO_SAMPLE = MODEL.config.num_frames
|
31 |
+
IMAGE_STATS = {"image_mean": [0.485, 0.456, 0.406], "image_std": [0.229, 0.224, 0.225]}
|
32 |
+
VAL_TRANSFORMS = Compose(
|
33 |
+
[
|
34 |
+
UniformTemporalSubsample(NUM_FRAMES_TO_SAMPLE),
|
35 |
+
Lambda(lambda x: x / 255.0),
|
36 |
+
Normalize(IMAGE_STATS["image_mean"], IMAGE_STATS["image_std"]),
|
37 |
+
Resize((RESIZE_TO, RESIZE_TO)),
|
38 |
+
]
|
39 |
+
)
|
40 |
+
LABELS = list(MODEL.config.label2id.keys())
|
41 |
+
|
42 |
+
|
43 |
+
def parse_video(video_file):
|
44 |
+
"""A utility to parse the input videos.
|
45 |
+
|
46 |
+
Reference: https://pyimagesearch.com/2018/11/12/yolo-object-detection-with-opencv/
|
47 |
+
"""
|
48 |
+
vs = cv2.VideoCapture(video_file)
|
49 |
+
|
50 |
+
# try to determine the total number of frames in the video file
|
51 |
+
try:
|
52 |
+
prop = (
|
53 |
+
cv2.cv.CV_CAP_PROP_FRAME_COUNT
|
54 |
+
if imutils.is_cv2()
|
55 |
+
else cv2.CAP_PROP_FRAME_COUNT
|
56 |
+
)
|
57 |
+
total = int(vs.get(prop))
|
58 |
+
print("[INFO] {} total frames in video".format(total))
|
59 |
+
|
60 |
+
# an error occurred while trying to determine the total
|
61 |
+
# number of frames in the video file
|
62 |
+
except:
|
63 |
+
print("[INFO] could not determine # of frames in video")
|
64 |
+
print("[INFO] no approx. completion time can be provided")
|
65 |
+
total = -1
|
66 |
+
|
67 |
+
frames = []
|
68 |
+
|
69 |
+
# loop over frames from the video file stream
|
70 |
+
while True:
|
71 |
+
# read the next frame from the file
|
72 |
+
(grabbed, frame) = vs.read()
|
73 |
+
if frame is not None:
|
74 |
+
frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
|
75 |
+
frames.append(frame)
|
76 |
+
# if the frame was not grabbed, then we have reached the end
|
77 |
+
# of the stream
|
78 |
+
if not grabbed:
|
79 |
+
break
|
80 |
+
|
81 |
+
return frames
|
82 |
+
|
83 |
+
|
84 |
+
def preprocess_video(frames: list):
|
85 |
+
"""Utility to apply preprocessing transformations to a video tensor."""
|
86 |
+
# Each frame in the `frames` list has the shape: (height, width, num_channels).
|
87 |
+
# Collated together the `frames` has the the shape: (num_frames, height, width, num_channels).
|
88 |
+
# So, after converting the `frames` list to a torch tensor, we permute the shape
|
89 |
+
# such that it becomes (num_channels, num_frames, height, width) to make
|
90 |
+
# the shape compatible with the preprocessing transformations. After applying the
|
91 |
+
# preprocessing chain, we permute the shape to (num_frames, num_channels, height, width)
|
92 |
+
# to make it compatible with the model. Finally, we add a batch dimension so that our video
|
93 |
+
# classification model can operate on it.
|
94 |
+
video_tensor = torch.tensor(np.array(frames).astype(frames[0].dtype))
|
95 |
+
video_tensor = video_tensor.permute(
|
96 |
+
3, 0, 1, 2
|
97 |
+
) # (num_channels, num_frames, height, width)
|
98 |
+
video_tensor_pp = VAL_TRANSFORMS(video_tensor)
|
99 |
+
video_tensor_pp = video_tensor_pp.permute(
|
100 |
+
1, 0, 2, 3
|
101 |
+
) # (num_frames, num_channels, height, width)
|
102 |
+
video_tensor_pp = video_tensor_pp.unsqueeze(0)
|
103 |
+
return video_tensor_pp.to(DEVICE)
|
104 |
+
|
105 |
+
|
106 |
+
def infer(video_file):
|
107 |
+
frames = parse_video(video_file)
|
108 |
+
video_tensor = preprocess_video(frames)
|
109 |
+
inputs = {"pixel_values": video_tensor}
|
110 |
+
|
111 |
+
# forward pass
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
112 |
with torch.no_grad():
|
113 |
+
outputs = MODEL(**inputs)
|
114 |
logits = outputs.logits
|
115 |
+
softmax_scores = torch.nn.functional.softmax(logits, dim=-1).squeeze(0)
|
116 |
+
confidences = {LABELS[i]: float(softmax_scores[i]) for i in range(len(LABELS))}
|
117 |
+
return confidences
|
118 |
+
|
119 |
+
|
120 |
+
gr.Interface(
|
121 |
+
fn=infer,
|
122 |
+
inputs=gr.Video(type="file"),
|
123 |
+
outputs=gr.Label(num_top_classes=3),
|
124 |
+
examples=[
|
125 |
+
["examples/babycrawling.mp4"],
|
126 |
+
["examples/baseball.mp4"],
|
127 |
+
["examples/balancebeam.mp4"],
|
128 |
+
],
|
129 |
+
title="VideoMAE fine-tuned on a subset of UCF-101",
|
130 |
+
description=(
|
131 |
+
"Gradio demo for VideoMAE for video classification. To use it, simply upload your video or click one of the"
|
132 |
+
" examples to load them. Read more at the links below."
|
133 |
+
),
|
134 |
+
article=(
|
135 |
+
"<div style='text-align: center;'><a href='https://huggingface.co/docs/transformers/model_doc/videomae' target='_blank'>VideoMAE</a>"
|
136 |
+
" <center><a href='https://huggingface.co/sayakpaul/videomae-base-finetuned-kinetics-finetuned-ucf101-subset' target='_blank'>Fine-tuned Model</a></center></div>"
|
137 |
+
),
|
138 |
+
allow_flagging=False,
|
139 |
+
allow_screenshot=False,
|
140 |
+
).launch()
|