Add application file

app.py

@@ -0,0 +1,224 @@
+from fastcore.all import *
+from fastai.vision.all import *
+
+learn = load_learner('./fast_ai_model_resnet18_new_labeling.pth')
+
+
+import cv2
+import torch
+import torchvision.transforms as tt
+import numpy as np
+# from facenet_pytorch import MTCNN
+from PIL import Image
+import numpy as np
+
+from retinaface import RetinaFace
+
+
+
+
+device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')
+
+def load_checkpoint(filepath):
+    checkpoint = torch.load(filepath, device)
+    model = checkpoint['model']
+    model.load_state_dict(checkpoint['state_dict'], strict=False)
+    for parameter in model.parameters():
+        parameter.requires_grad = False
+
+    model.eval()
+    return model
+
+
+# filepath = './models/9.pth'
+# loaded_model = load_checkpoint(filepath)
+
+loaded_model = learn
+
+class FaceDetector(object):
+    """
+    Face detector class
+    """
+
+    def __init__(self, detector,loaded_model,image=None):
+        self.detector = detector
+        self.loaded_model=loaded_model
+        self.image = image
+
+    def _draw(self, frame, boxes, probs, landmarks):
+        """
+        Draw landmarks and boxes for each face detected
+        """
+        try:
+            for box, prob, ld in zip(boxes, probs, landmarks):
+                # Draw rectangle on frame
+                box = box.astype('int')
+                ld = ld.astype('int')
+                cv2.rectangle(frame,
+                              (box[0], box[1]),
+                              (box[2], box[3]),
+                              (0, 0, 255),
+                              thickness=2)
+
+                # Show probability
+                cv2.putText(frame, str(
+                    prob), (box[2], box[3]), cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 0, 255), 2, cv2.LINE_AA)
+
+                # Draw landmarks
+                cv2.circle(frame, tuple(ld[0]), 5, (0, 0, 255), -1)
+                cv2.circle(frame, tuple(ld[1]), 5, (0, 0, 255), -1)
+                cv2.circle(frame, tuple(ld[2]), 5, (0, 0, 255), -1)
+                cv2.circle(frame, tuple(ld[3]), 5, (0, 0, 255), -1)
+                cv2.circle(frame, tuple(ld[4]), 5, (0, 0, 255), -1)
+        except Exception as e:
+            # print(e)
+            pass
+
+        return frame
+
+    # def _capture(self,frame,boxes,probs,landmarks):
+    #     sampleNum = 0
+    #     while True:   
+    #         try:
+    #             sampleNum = sampleNum+1
+    #             boxes = boxes.astype('int')
+    #             face = frame[boxes[0,1]:boxes[0,3],boxes[0,0]:boxes[0,2]]
+    #             cv2.imwrite(str(sampleNum) + ".jpg",face)                    
+    #         except Exception as e:
+    #             # print(e)
+    #             pass
+    #         if sampleNum >2:
+    #             break
+
+
+    def _recognize(self,loaded_model,frame,boxes):
+        try:
+            
+            for box in boxes:
+                # box = box.astype('int')
+                # face = frame[box[0][1]:box[0][3],box[0][0]:box[0][2]]
+                face = frame[box[1]:box[3],box[0]:box[2]]
+                # pil_image = Image.fromarray(face, mode="RGB")
+
+                labels,_,probs = loaded_model.predict(face)
+                # label= 1
+
+                
+                
+
+                label = labels if (probs[np.argmax(probs)] > 0.8) else "Unknown"
+                probs = probs[np.argmax(probs)] if (probs[np.argmax(probs)] > 0.8) else np.nan
+
+                # if prediction == 0:
+                #     label = "Fadli"
+                # elif prediction == 1:
+                #     label = "Aziz"
+                # elif prediction == 2:
+                #     label = "Eka"
+                # else:
+                #     label = "Unknown"
+                _ = "" if probs == np.nan else ": {:.2f}".format(float(probs))
+
+                cv2.putText(frame, label + _ , (box[2], box[3]), cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 0, 255), 2, cv2.LINE_AA)
+
+                # cv2.putText(frame,str(["{:.2f}".format(x)for x in probs]), (box[2]-100, box[3]+30), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 0, 255), 2, cv2.LINE_AA)
+
+                # cv2.putText(frame,"{:.2f}".format(float(probs[np.argmax(probs)])), (box[2]-100, box[3]+30), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 0, 255), 2, cv2.LINE_AA)
+
+
+
+
+                cv2.rectangle(frame,
+                              (box[0], box[1]),
+                              (box[2], box[3]),
+                              (0, 0, 255),
+                              thickness=2)
+                # cv2.putText(frame, str(
+                #     prediction), (box[0][2], box[0][3]), cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 0, 255), 2, cv2.LINE_AA)
+            return frame
+        except Exception as e:
+            print(e,'eaea')
+            pass
+    
+    def run_on_image(self,image):
+        frame = image
+        try:
+            # detect face box, probability and landmarks
+            json_face = self.detector.detect_faces(frame)
+
+            # boxes = [json_face['face_{}'.format(x)]['facial_area'] for x in range(1,len(json_face)+1)]
+            # probs = ["{:.2f}".format(json_face['face_{}'.format(x)]['score']) for x in range(1,len(json_face)+1)]
+            # landmarks = [json_face['face_{}'.format(x)]['landmarks'] for x in range(1,len(json_face)+1)]
+
+            boxes = []
+            probs = []
+            landmarks = []
+            for key, value in json_face.items():
+                boxes.append(value['facial_area'])
+                probs.append("{:.2f}".format(value['score']))
+                landmarks.append(value['landmarks'])
+            
+            return Image.fromarray(self._recognize(loaded_model,frame,boxes))
+            # self._draw(frame, boxes, probs, landmarks)
+            
+        except Exception as e:
+            print(e)
+            pass
+
+
+    def run(self):
+        """
+            Run the FaceDetector and draw landmarks and boxes around detected faces
+        """
+        cap = cv2.VideoCapture(0)
+
+        while True:
+            ret, frame = cap.read()
+            try:
+                # detect face box, probability and landmarks
+                json_face = self.detector.detect_faces(frame)
+
+                # boxes = [json_face['face_{}'.format(x)]['facial_area'] for x in range(1,len(json_face)+1)]
+                # probs = ["{:.2f}".format(json_face['face_{}'.format(x)]['score']) for x in range(1,len(json_face)+1)]
+                # landmarks = [json_face['face_{}'.format(x)]['landmarks'] for x in range(1,len(json_face)+1)]
+
+                boxes = []
+                probs = []
+                landmarks = []
+                for key, value in json_face.items():
+                    boxes.append(value['facial_area'])
+                    probs.append("{:.2f}".format(value['score']))
+                    landmarks.append(value['landmarks'])
+
+                self._recognize(loaded_model,frame,boxes)
+                # self._draw(frame, boxes, probs, landmarks)
+                
+            except Exception as e:
+                print(e)
+                pass
+
+            # Show the frame
+            cv2.imshow('Face Detection', frame)
+
+            if cv2.waitKey(1) & 0xFF == ord('q'):
+                break
+
+        cap.release()
+        cv2.destroyAllWindows()
+        
+        
+# Run the app
+
+# fcd = FaceDetector(RetinaFace,loaded_model)
+detector = RetinaFace
+# fcd = FaceDetector(detector,loaded_model)
+# fcd.run()
+
+fcd = FaceDetector(detector,loaded_model)
+
+import gradio as gr
+
+gr.Interface(fn=fcd.run_on_image, 
+             inputs=gr.Image(),
+             outputs=gr.Image(),
+             ).launch()

fast_ai_model_resnet18_new_labeling.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6799a295f99c4fe0a40e75ca174a7516ce6ea912228a9056ceee5a5683d5b818
+size 46969633

requirements.txt

@@ -0,0 +1,11 @@
+fastai==2.7.8
+fastapi==0.85.1
+fastcore==1.5.27
+fastdownload==0.0.7
+fastprogress==1.0.3
+retina-face==0.0.12
+torch==1.12.1
+torchaudio==0.12.1
+torchvision==0.13.1
+numpy==1.23.2
+Pillow==9.2.0