test

app.py

@@ -0,0 +1,35 @@
+import gradio as gr
+import cv2
+import numpy as np
+import torch
+import torch.nn as nn
+import torchvision.models as models
+import einops
+
+def predict(img):
+    device = 'cuda' if torch.cuda.is_available() else 'cpu'
+    model = models.resnet50()
+    model.fc = nn.Linear(2048, 720)
+    resume_path = 'full+++++.pth'
+    model.load_state_dict(torch.load(resume_path))
+    model.to(device)
+    with torch.no_grad():
+        model.eval()
+        img = cv2.resize(img, (224, 224))/255.
+        img = np.stack([einops.rearrange(img, 'h w c -> c h w')], 0)
+        img = torch.Tensor(img).float().to(device)
+        pred = model(img)  
+        max_pred = torch.argsort(pred, dim=1, descending=True)
+        max_h = (max_pred[0][0] // 60).item()
+        max_m = (max_pred[0][0] % 60).item()
+    return '{}:{}'.format(str(max_h), str(max_m).zfill(2))
+
+inputs = gr.inputs.Image()
+
+io = gr.Interface(
+    fn=predict,
+    inputs=inputs,
+    outputs="text",
+)
+
+io.launch(share=True)

full+++++.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7f709d84fc468882e84007a810253c02933154988009016fe0cb76a5cc99e5b6
+size 100250049

requirements.txt

@@ -0,0 +1,6 @@
+cv2
+numpy
+torch
+torchvision
+einops
+gradio