FER Working on RPi with ~30sec predict time

FER/detectfaces.py

@@ -6,8 +6,10 @@ import time
 from PIL import Image
 from main import RecorderMeter1, RecorderMeter  # noqa: F401
 
-# Define the path to the model checkpoint
-model_path = os.path.abspath(r"FER\models\checkpoints\raf-db-model_best.pth")
+script_dir = os.path.dirname(os.path.abspath(__file__))
+
+# Construct the full path to the model file
+model_path = os.path.join(script_dir,"models","checkpoints","raf-db-model_best.pth")
 
 # Determine the available device for model execution
 if torch.backends.mps.is_available():
@@ -86,9 +88,6 @@ def imagecapture(model):
             cv2.data.haarcascades + "haarcascade_frontalface_default.xml"
         ).detectMultiScale(gray, scaleFactor=1.3, minNeighbors=5, minSize=(30, 30))
 
-        # Display the current frame
-        cv2.imshow("Webcam", frame)
-
         # If faces are detected, proceed with prediction
         if len(faces) > 0:
             currtimeimg = time.strftime("%H:%M:%S")
@@ -107,7 +106,7 @@ def imagecapture(model):
             starttime = time.strftime("%H:%M:%S")
             print(f"-->Prediction starting at {starttime}")
             # Perform emotion prediction
-            predict(model, image_path=face_pil_image)
+            face_detection(model, image_path=face_pil_image)
             # Record the prediction end time
             endtime = time.strftime("%H:%M:%S")
             print(f"-->Done prediction at {endtime}")

FER/models/PosterV2_7cls.py

@@ -309,7 +309,7 @@ class pyramid_trans_expr2(nn.Module):
         mobilefacenet_path = os.path.join(
             script_dir, "pretrain", "mobilefacenet_model_best.pth.tar"
         )
-        ir50_path = os.path.join(script_dir, r"pretrain\ir50.pth")
+        ir50_path = os.path.join(script_dir, "pretrain","ir50.pth")
 
         print(mobilefacenet_path)
         face_landback_checkpoint = torch.load(

FER/models/checkpoints/raf-db-model_best.pth

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:d9bf1d0d88238966ce0d1a289a2bb5f927ec2fe635ef1ec4396c323028924701
-size 238971279

FER/prediction.py

@@ -23,7 +23,7 @@ model = model.to(device)
 script_dir = os.path.dirname(os.path.abspath(__file__))
 
 # Construct the full path to the model file
-model_path = os.path.join(script_dir, r"models\checkpoints\raf-db-model_best.pth")
+model_path = os.path.join(script_dir,"models","checkpoints","raf-db-model_best.pth")
 
 image_arr = []
 for foldername, subfolders, filenames in os.walk("../FER/Images/"):
@@ -32,27 +32,6 @@ for foldername, subfolders, filenames in os.walk("../FER/Images/"):
         file_path = os.path.join(foldername, filename)
         image_arr.append(f"{file_path}")
 
-
-def main():
-    if model_path is not None:
-        if os.path.isfile(model_path):
-            print("=> loading checkpoint '{}'".format(model_path))
-            checkpoint = torch.load(model_path, map_location=device)
-            best_acc = checkpoint["best_acc"]
-            best_acc = best_acc.to()
-            print(f"best_acc:{best_acc}")
-            model.load_state_dict(checkpoint["state_dict"])
-            print(
-                "=> loaded checkpoint '{}' (epoch {})".format(
-                    model_path, checkpoint["epoch"]
-                )
-            )
-        else:
-            print("[!] prediction.py => no checkpoint found at '{}'".format(model_path))
-        predict(model, image_path=image_arr)
-        return
-
-
 def predict(model, image_path):
     from face_detection import face_detection
 
@@ -102,7 +81,3 @@ def predict(model, image_path):
             f"-->Image Path {image_path} [!] The predicted labels are {y_pred} and the label is {labels}"
         )
     return
-
-
-if __name__ == "__main__":
-    main()

camera.py

@@ -0,0 +1,15 @@
+import cv2
+
+# initialize the camera
+cap = cv2.VideoCapture(0, cv2.CAP_V4L2)
+
+# capture a frame
+ret, frame = cap.read()
+
+# release the camera
+cap.release()
+
+# save the captured frame
+cv2.imwrite('captured_image.jpg', frame)
+
+

pyproject.toml

@@ -134,3 +134,6 @@ dependencies = [
 
 [[tool.uv.index]]
 url = "https://download.pytorch.org/whl/cu124"
+
+[[tool.uv.index]]
+url = "https://pypi.org/simple"