search

bearsprogrammer / real-time-deep-face-recognition

using facenet algorithm
MIT License
233 stars 100 forks source link

the classification is same when detect two person #6

Open billtiger opened 7 years ago

billtiger commented 7 years ago

when I have two persons to detect, it could detect two person! but the classification is also the same when detect two person,e.g.two person's name are bill and lisa,the classification result is two person are bill or two person are lisa,what's the problem?your reply for my problem will be highly appreciated!thank you! @bearsprogrammer

so-as commented 7 years ago

when I run this project to do face detect with tensorflow 1.2 version on the ubuntu 14.04, I met the error: 2017-10-13 00:39:02.833148: I tensorflow/core/common_runtime/gpu/gpu_device.cc:965] Found device 0 with properties: name: Tesla M40 major: 5 minor: 2 memoryClockRate(GHz): 1.112 pciBusID: 0000:82:00.0 totalMemory: 11.93GiB freeMemory: 11.82GiB 2017-10-13 00:39:02.833193: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1055] Creating TensorFlow device (/device:GPU:0) -> (device: 0, name: Tesla M40, pci bus id: 0000:82:00.0, compute capability: 5.2) 2017-10-13 00:39:02.863038: E tensorflow/stream_executor/cuda/cuda_driver.cc:936] failed to allocate 11.93G (12808486912 bytes) from device: CUDA_ERROR_OUT_OF_MEMORY It seem the 12G memory is not enough for mtcnn model. Could you tell me which GPU device you use, how much memory does the mtcnn detection algorith need? @billtiger

samrid commented 7 years ago

@so-as Change "per_process_gpu_memory_fraction=[x]" to "per_process_gpu_memory_fraction=0.4"

samrid commented 7 years ago

I have the same issue @billtiger It's can solved by

billtiger commented 7 years ago

thanks,you solved my problem!it worked! @samrid

billtiger commented 7 years ago

i can run mtcnn on my i5 CPU computer,so it doesn't need much memory,you can check there are other reason cause this error? @so-as

GarvLodha commented 6 years ago

@billtiger and @samrid guys, i am also having same problem as @billtiger. I did the changes as suggested by @samrid. But, still having the same classification issue. Can you guys please have a look at my code. Thanks!!

realtime_facenet_git.txt

OR

cropped.append(frame[bb[i][1]:bb[i][3], bb[i][0]:bb[i][2], :])

cropped[0] = facenet.flip(cropped[0], False)

                    cropped[i] = facenet.flip(cropped[i], False)
                    ##scaled.append(misc.imresize(cropped[0], (image_size, image_size), interp='bilinear'))
                    scaled.append(misc.imresize(cropped[i], (image_size, image_size), interp='bilinear'))
                    ##scaled[0] = cv2.resize(scaled[0], (input_image_size,input_image_size),interpolation=cv2.INTER_CUBIC)
                    scaled[i] = cv2.resize(scaled[i], (input_image_size,input_image_size),interpolation=cv2.INTER_CUBIC)
                    ##scaled[0] = facenet.prewhiten(scaled[0])
                    scaled[i] = facenet.prewhiten(scaled[i])
                    ##scaled_reshape.append(scaled[0].reshape(-1,input_image_size,input_image_size,3))
                    scaled_reshape.append(scaled[i].reshape(-1,input_image_size,input_image_size,3))
                    ##feed_dict = {images_placeholder: scaled_reshape[0], phase_train_placeholder: False}
                    feed_dict = {images_placeholder: scaled_reshape[i], phase_train_placeholder: False}
                    emb_array[0, :] = sess.run(embeddings, feed_dict=feed_dict)
                    predictions = model.predict_proba(emb_array)
                    best_class_indices = np.argmax(predictions, axis=1)
                    best_class_probabilities = predictions[np.arange(len(best_class_indices)), best_class_indices]
                    cv2.rectangle(frame, (bb[i][0], bb[i][1]), (bb[i][2], bb[i][3]), (0, 255, 0), 2)    #boxing face
dwikysyahbana commented 6 years ago

same with @GarvLodha i still have the same problemm

GarvLodha commented 6 years ago

@dwikysyahbana , try this:

j = 0 #appended for i in range(nrof_faces): emb_array = np.zeros((1, embedding_size))

                    bb[i][0] = det[i][0]
                    bb[i][1] = det[i][1]
                    bb[i][2] = det[i][2]
                    bb[i][3] = det[i][3]

                    # inner exception
                    if bb[i][0] <= 0 or bb[i][1] <= 0 or bb[i][2] >= len(frame[0]) or bb[i][3] >= len(frame):
                        print('face is inner of range!')
                        continue

                    cropped.append(frame[bb[i][1]:bb[i][3], bb[i][0]:bb[i][2], :])
                    ##cropped[0] = facenet.flip(cropped[0], False)
                    ##cropped[i] = facenet.flip(cropped[i], False)
                    cropped[j] = facenet.flip(cropped[j], False)
                    ##scaled.append(misc.imresize(cropped[0], (image_size, image_size), interp='bilinear'))
                    ##scaled.append(misc.imresize(cropped[i], (image_size, image_size), interp='bilinear'))
                    scaled.append(misc.imresize(cropped[j], (image_size, image_size), interp='bilinear'))
                    ##scaled[0] = cv2.resize(scaled[0], (input_image_size,input_image_size),interpolation=cv2.INTER_CUBIC)
                    ##scaled[i] = cv2.resize(scaled[i], (input_image_size,input_image_size),interpolation=cv2.INTER_CUBIC)
                    scaled[j] = cv2.resize(scaled[j], (input_image_size,input_image_size),interpolation=cv2.INTER_CUBIC)
                    ##scaled[0] = facenet.prewhiten(scaled[0])
                    ##scaled[i] = facenet.prewhiten(scaled[i])
                    scaled[j] = facenet.prewhiten(scaled[j])
                    ##scaled_reshape.append(scaled[0].reshape(-1,input_image_size,input_image_size,3))
                    ##scaled_reshape.append(scaled[i].reshape(-1,input_image_size,input_image_size,3))
                    scaled_reshape.append(scaled[j].reshape(-1,input_image_size,input_image_size,3))
                    ##feed_dict = {images_placeholder: scaled_reshape[0], phase_train_placeholder: False}
                    ##feed_dict = {images_placeholder: scaled_reshape[i], phase_train_placeholder: False}
                    feed_dict = {images_placeholder: scaled_reshape[j], phase_train_placeholder: False}
                    emb_array[0, :] = sess.run(embeddings, feed_dict=feed_dict)
                    predictions = model.predict_proba(emb_array)
                    best_class_indices = np.argmax(predictions, axis=1)
                    best_class_probabilities = predictions[np.arange(len(best_class_indices)), best_class_indices]
                    cv2.rectangle(frame, (bb[i][0], bb[i][1]), (bb[i][2], bb[i][3]), (0, 255, 0), 2)    #boxing face

                    #plot result idx under box
                    text_x = bb[i][0]
                     text_y = bb[i][3] + 20
                    ##print('result: ', best_class_indices[0])
                    print('result: ', best_class_indices[0])
                    print('prob: ', best_class_probabilities[0])
                    if best_class_probabilities > 0.40:
                         for H_i in HumanNames:
                              ##if HumanNames[best_class_indices[0]] == H_i:
                              if HumanNames[best_class_indices[0]-1] == H_i:
                                   ##result_names = HumanNames[best_class_indices[0]]
                                   result_names = HumanNames[best_class_indices[0]-1]
                                   cv2.putText(frame, result_names, (text_x, text_y), cv2.FONT_HERSHEY_COMPLEX_SMALL,1, (0, 0, 255), thickness=1, lineType=2)
                    else:
                     cv2.putText(frame,'unknown', (text_x, text_y), cv2.FONT_HERSHEY_COMPLEX_SMALL,1, (0, 0, 255), thickness=1, lineType=2)
                j += 1
            else:
                print('Unable to align')
Gabit07 commented 5 years ago

I have the same issue @billtiger It's can solved by

  • Change "cropped[0]" to "cropped[i]"
  • Change "scaled[0]" to "scaled[i]"
  • Change "scaled_reshape[0]" to "scaled_reshape[i]" it's worked for me. thank @bearsprogrammer for your shared.

I also still having the same issue

Gabit07 commented 5 years ago

@billtiger and @samrid guys, i am also having same problem as @billtiger. I did the changes as suggested by @samrid. But, still having the same classification issue. Can you guys please have a look at my code. Thanks!!

realtime_facenet_git.txt

OR

cropped.append(frame[bb[i][1]:bb[i][3], bb[i][0]:bb[i][2], :])

cropped[0] = facenet.flip(cropped[0], False)

cropped[i] = facenet.flip(cropped[i], False)

scaled.append(misc.imresize(cropped[0], (image_size, image_size), interp='bilinear'))

scaled.append(misc.imresize(cropped[i], (image_size, image_size), interp='bilinear'))

scaled[0] = cv2.resize(scaled[0], (input_image_size,input_image_size),interpolation=cv2.INTER_CUBIC)

scaled[i] = cv2.resize(scaled[i], (input_image_size,input_image_size),interpolation=cv2.INTER_CUBIC)

scaled[0] = facenet.prewhiten(scaled[0])

scaled[i] = facenet.prewhiten(scaled[i])

scaled_reshape.append(scaled[0].reshape(-1,input_image_size,input_image_size,3))

scaled_reshape.append(scaled[i].reshape(-1,input_image_size,input_image_size,3))

feed_dict = {images_placeholder: scaled_reshape[0], phase_train_placeholder: False}

feed_dict = {images_placeholder: scaled_reshape[i], phase_train_placeholder: False} emb_array[0, :] = sess.run(embeddings, feed_dict=feed_dict) predictions = model.predict_proba(emb_array) best_class_indices = np.argmax(predictions, axis=1) best_class_probabilities = predictions[np.arange(len(best_class_indices)), best_class_indices] cv2.rectangle(frame, (bb[i][0], bb[i][1]), (bb[i][2], bb[i][3]), (0, 255, 0), 2) #boxing face

Yesterday I also had the same issue. Solved it by fixing path. In the Make_classifier_git.py file 1st path has to be the folder of aligned images. I mistakenly did path to the original image folder, not to the aligned image folder. I hope this solves your problem

Gabit07 commented 5 years ago

same with @GarvLodha i still have the same problemm

I just suggest an solution in the above