Update app.py
Browse files
app.py
CHANGED
@@ -1,3 +1,5 @@
|
|
|
|
|
|
1 |
"""
|
2 |
=====================================================
|
3 |
Optical Flow: Predicting movement with the RAFT model
|
@@ -74,10 +76,8 @@ _ = urlretrieve(video_url, video_path)
|
|
74 |
from torchvision.io import read_video
|
75 |
frames, _, _ = read_video(str(video_path), output_format="TCHW")
|
76 |
|
77 |
-
|
78 |
-
|
79 |
-
|
80 |
-
plot(img1_batch)
|
81 |
|
82 |
#########################
|
83 |
# The RAFT model accepts RGB images. We first get the frames from
|
@@ -92,15 +92,15 @@ weights = Raft_Large_Weights.DEFAULT
|
|
92 |
transforms = weights.transforms()
|
93 |
|
94 |
|
95 |
-
def preprocess(
|
96 |
-
|
97 |
-
|
98 |
-
return transforms(
|
99 |
|
100 |
|
101 |
-
|
102 |
|
103 |
-
print(f"shape = {
|
104 |
|
105 |
|
106 |
####################################
|
@@ -120,7 +120,7 @@ device = "cuda" if torch.cuda.is_available() else "cpu"
|
|
120 |
model = raft_large(weights=Raft_Large_Weights.DEFAULT, progress=False).to(device)
|
121 |
model = model.eval()
|
122 |
|
123 |
-
list_of_flows = model(
|
124 |
print(f"type = {type(list_of_flows)}")
|
125 |
print(f"length = {len(list_of_flows)} = number of iterations of the model")
|
126 |
|
@@ -160,9 +160,9 @@ from torchvision.utils import flow_to_image
|
|
160 |
flow_imgs = flow_to_image(predicted_flows)
|
161 |
|
162 |
# The images have been mapped into [-1, 1] but for plotting we want them in [0, 1]
|
163 |
-
|
164 |
|
165 |
-
grid = [[img1, flow_img] for (img1, flow_img) in zip(
|
166 |
plot(grid)
|
167 |
|
168 |
####################################
|
|
|
1 |
+
import gradio as gr
|
2 |
+
|
3 |
"""
|
4 |
=====================================================
|
5 |
Optical Flow: Predicting movement with the RAFT model
|
|
|
76 |
from torchvision.io import read_video
|
77 |
frames, _, _ = read_video(str(video_path), output_format="TCHW")
|
78 |
|
79 |
+
img1= [frames[100]
|
80 |
+
img2 = [frames[101]
|
|
|
|
|
81 |
|
82 |
#########################
|
83 |
# The RAFT model accepts RGB images. We first get the frames from
|
|
|
92 |
transforms = weights.transforms()
|
93 |
|
94 |
|
95 |
+
def preprocess(img, img2):
|
96 |
+
img1 = F.resize(img1, size=[520, 960])
|
97 |
+
img2 = F.resize(img2, size=[520, 960])
|
98 |
+
return transforms(img1, img2)
|
99 |
|
100 |
|
101 |
+
img1, img2 = preprocess(img1, img2)
|
102 |
|
103 |
+
print(f"shape = {img1.shape}, dtype = {img1.dtype}")
|
104 |
|
105 |
|
106 |
####################################
|
|
|
120 |
model = raft_large(weights=Raft_Large_Weights.DEFAULT, progress=False).to(device)
|
121 |
model = model.eval()
|
122 |
|
123 |
+
list_of_flows = model(img1.to(device), img2.to(device))
|
124 |
print(f"type = {type(list_of_flows)}")
|
125 |
print(f"length = {len(list_of_flows)} = number of iterations of the model")
|
126 |
|
|
|
160 |
flow_imgs = flow_to_image(predicted_flows)
|
161 |
|
162 |
# The images have been mapped into [-1, 1] but for plotting we want them in [0, 1]
|
163 |
+
img1 = [(img1 + 1) / 2 for img1 in img1]
|
164 |
|
165 |
+
grid = [[img1, flow_img] for (img1, flow_img) in zip(img1, flow_imgs)]
|
166 |
plot(grid)
|
167 |
|
168 |
####################################
|