create file .pickle

[phamquangnam]

Hi .. How can I create file .pickle for other video demo? Please help me

Thanks!

[zhangdengp]

i meet the same problem

[leijuzi]

do you solve it?

[InzamamAnwar]

@phamquangnam you can use any off-the-shelve person/pedestrian detector, Faster R-CNN/Yolo-v3/SSD etc to generate the detections

[Knightfire1998]

hello can you tell me how to save them in the required format as i am unable to get the right format i am using tensorflow object detection model for detections please help

[InzamamAnwar]

@Knightfire1998 the format is as follows: From each frame there will be array of array containing key-points for each person in the following format

[top_right_x, top_right_y, bottom_left_x, bottom_left_y, confidence]

For example in frame 0 there may be 2 person so the detection size will be (2, 5) and in frame 10 there may be 1 person so the detection size will be (1, 5).

I have used faster_rcnn_inception_v2_coco_2018_01_28 from tensorflow object detection. You can use this model if working with person/pedestrian class.

[Knightfire1998]

@InzamamAnwar can you share code snippet when you pickle it as I m using the box object to save the Co-ordinates their array of size(1,5) only please help me with it thank you or email me if possible reshikeshdhanrale@gmail.com

[Knightfire1998]

while True:
    r, img = cap.read()
    if img is not None:
        img = cv2.resize(img, (1280, 720))
        #print(img)
        boxes, scores, classes, num = odapi.processFrame(img)

    else:
        break

    # Visualization of the results of a detection.

    for i in range(len(boxes)):
        # Class 1 represents human
        if classes[i] == 1 and scores[i] > threshold:

            box = boxes[i]
            a=[box[0],box[1],box[2],box[3],0]
            li.append(a)

            print("*****")

            cv2.rectangle(img,(box[1],box[0]),(box[3],box[2]),(255,0,0),2)

        if len(li)!=0:
            ma.append(np.array(li[:], ndmin=2))

        print("-----")
        li=[]
    cv2.imshow("preview", img)
    pickle_out = open("1.pickle","wb")

    pickle.dump(ma[:], pickle_out)
    pickle_out.close()
    key = cv2.waitKey(1)
    if key & 0xFF == ord('q'):
        break
object1 = pd.read_pickle(r'1.pickle')
print(object1)

[Knightfire1998]

see the code and tell me whats wrong plz as after giving the pickle file 1.pickle to master.py as input it shows problem wit concat in embeddings as index[0] has dim 1,5 and index[1] has 1,6

[InzamamAnwar]

@Knightfire1998 I have written the while loop and tested with the data provided with this repo. Insert this loop instead of yours and It will work. If not, let me know

det = []
while True:
    r, img = cap.read()
    if r:
        out = []
        boxes, scores, classes, num = odapi.processFrame(img)
        for i in range(len(boxes)):
            if classes[i] == 1 and scores[i] > threshold:
                box = boxes[i]
                out.append([box[1], box[0], box[3], box[2], scores[i]])
        if len(out) == 0:
            det.append(np.empty(shape=(0, 5), dtype=np.float32))
        else:
            det.append(np.asanyarray(out, dtype=np.float32))
    else:
        break

[Knightfire1998]

@InzamamAnwar i m getting the following error after running the loop i did pickle.dump(det, pickle_out) then a repllaced 4p-c4 with my own video file and pickle file with same name i got the following error:

frame_id: 100 Traceback (most recent call last): File "master.py", line 291, in demo(save=True) File "master.py", line 220, in demo embedding = get_embedding(boxes, frames, box_to_vect, sess) File "master.py", line 94, in get_embedding embedding = box_to_vect.inference(image_batch, bbox_batch, box_ind_batch, sess) File "F:\study material\AI\Project\Multi-Camera-Object-Tracking-via-Transferring-Representation-to-Top-View-master\box2vec\resnet.py", line 167, in inference embeddings = sess.run(self.embeddings, feed_dict=feed_dict) File "C:\Users\Reshikesh\Anaconda3\envs\test2\lib\site-packages\tensorflow_core\python\client\session.py", line 956, in run run_metadata_ptr) File "C:\Users\Reshikesh\Anaconda3\envs\test2\lib\site-packages\tensorflow_core\python\client\session.py", line 1156, in _run (np_val.shape, subfeed_t.name, str(subfeed_t.get_shape()))) ValueError: Cannot feed value of shape (6, 6) for Tensor 'Placeholder_1:0', which has shape '(?, 5)'

[InzamamAnwar]

Have you used this type of snippet for pickle dumping.

with open("filename.pickle", "wb") as f:
    pickle.dump(det, f)

[Knightfire1998]

pickle_out = open("1.pickle","wb") pickle.dump(det, pickle_out) pickle_out.close()
i used this

[Knightfire1998]

wont it work? should i change it?

[InzamamAnwar]

I think it should work. Can you please use 4p-c4.avi and generate detections from this and than run the code. If possible you can upload your code over github and I will look into it

[Knightfire1998]

okay i m running that on the file 4p-c4 i guess its problem with the number of maximum entities i m just checking that i ll upload the code in a while Thanks for helping man @InzamamAnwar

[Knightfire1998]

i used 4p-c4 but got this error

frame_id: 157 Traceback (most recent call last): File "master.py", line 291, in demo(save=True) File "master.py", line 220, in demo embedding = get_embedding(boxes, frames, box_to_vect, sess) File "master.py", line 94, in get_embedding embedding = box_to_vect.inference(image_batch, bbox_batch, box_ind_batch, sess) File "F:\study material\AI\Project\Multi-Camera-Object-Tracking-via-Transferring-Representation-to-Top-View-master\box2vec\resnet.py", line 167, in inference embeddings = sess.run(self.embeddings, feed_dict=feed_dict) File "C:\Users\Reshikesh\Anaconda3\envs\test2\lib\site-packages\tensorflow_core\python\client\session.py", line 956, in run run_metadata_ptr) File "C:\Users\Reshikesh\Anaconda3\envs\test2\lib\site-packages\tensorflow_core\python\client\session.py", line 1156, in _run (np_val.shape, subfeed_t.name, str(subfeed_t.get_shape()))) ValueError: Cannot feed value of shape (1, 6) for Tensor 'Placeholder_1:0', which has shape '(?, 5)'

[Knightfire1998]

i have uploaded the code on my repo as object detector with surveillance system plz take a look master.py is as it is no major changes have been made

[Knightfire1998]

@InzamamAnwar

[leconghieu99]

@Knightfire1998 Can you email me the full version of this section? thank you hieus.lecong@gmail.com