Logging Object Detection Results of Video in the Excel File

[Aar-Kay]

Hi,

I am new to coding and posting first time on GitHub.

Well, I was able to successfully train the model on the custom dataset and I am using it to detect some signs in the videos.
I can see the results in the video as well as in the terminal.

Now I want my results to be saved in an excel file which should have a log of all the objects detected along with the time of the video.

Can someone please suggest how it can be done?

[glenn-jocher]

@Aar-Kay hey buddy, congrats on your first training! You can produce detection logs as text files from video frames like this:

python detect.py --source video.mp4 --save-txt

The part of detect.py that saves these text files is here, you could introduce your own code to update an excel file instead:

yolov5/detect.py

Lines 103 to 110 in 73a0669

	# Write results
	for *xyxy, conf, cls in reversed(det):
	if save_txt: # Write to file
	xywh = (xyxy2xywh(torch.tensor(xyxy).view(1, 4)) / gn).view(-1).tolist() # normalized xywh
	line = (cls, *xywh, conf) if opt.save_conf else (cls, *xywh) # label format
	with open(txt_path + '.txt', 'a') as f:
	f.write(('%g ' * len(line)).rstrip() % line + '\n')

[mintu07ruet]

Thank you so much @Aar-Kay for sharing the code. I will go through in it. Appreciate!

[mintu07ruet]

Hi Aar-Kay
I am getting the following error
ImportError: cannot import name 'plot_one_box' from 'utils.plots' (C:\Users\mxm5116\yolov5\utils\plots.py)
any idea what can solve this issue? Also may I get your email, so that I can contact with through email. Thanks!

[mintu07ruet]

Hi Aar-kay,
There was mismatch with version which produced the error. your modified detect.py file is running, but could you please mention, how can I get the time stamp as in the log file, it is recording Frame, but I did not see any time stamp. Thanks

[renjirow]

can u share to us the file of ur project

[Asphyxcome]

hi, thanks for sharing the code, but when I tried to run the code, I got a few errors. Could you send the whole file where you made the changes and tell me how to output the data from yolo to excel file?

[mintu07ruet]

glenn-jocher@ can detect.py file can read video file directly from outlook one drive?

[glenn-jocher]

@mintu07ruet yes detect.py can load remote URL sources, i.e.:

!python detect.py --weights yolov5s.pt --source https://ultralytics.com/assets/decelera_landscape.mov  # video URL

[glenn-jocher]

[mintu07ruet]

glenn-jocher@ I am getting the above error when I am trying to read the video file from outlook onedrive. Any suggestions to run it. TIA!

[glenn-jocher]

@mintu07ruet if you have a complex URL you should always enclose it in quotation marks --source 'URL'

[parmarth-1208]

hey i just wanted to ask in the below code what path will replace "yolov8.weights" and "yolov8.cfg" ?
import cv2
import numpy as np
from itertools import combinations
import openpyxl

Function to calculate center point of a bounding box

def calculate_center(bbox):
return ((bbox[0] + bbox[2]) / 2, (bbox[1] + bbox[3]) / 2)

Function to compute Euclidean distance between two points

def euclidean_distance(point1, point2):
return np.sqrt((point1[0] - point2[0])**2 + (point1[1] - point2[1])**2)

Function to group characters based on proximity

def group_characters(bboxes):
center_points = [calculate_center(bbox) for bbox in bboxes]
distance_matrix = np.zeros((len(bboxes), len(bboxes)))
for i, center1 in enumerate(center_points):
for j, center2 in enumerate(center_points):
distance_matrix[i, j] = euclidean_distance(center1, center2)

proximity_threshold = 50
groups = []
for i, j in combinations(range(len(bboxes)), 2):
    if distance_matrix[i, j] < proximity_threshold:
        group_found = False
        for group in groups:
            if i in group or j in group:
                group.add(i)
                group.add(j)
                group_found = True
                break
        if not group_found:
            groups.append({i, j})
return groups

Load the pre-trained YOLOv8 model

net = cv2.dnn.readNet("yolov8.weights", "yolov8.cfg")
classes = ["0", "1", "2", "3", "4", "5", "6", "7", "8", "9",
"A", "B", "C", "D", "E", "F", "G", "H", "I", "J",
"K", "L", "M", "N", "O", "P", "Q", "R", "S", "T",
"U", "V", "W", "X", "Y", "Z"]
output_layers = net.getUnconnectedOutLayersNames()

Create Excel workbook and worksheet

wb = openpyxl.Workbook()
ws = wb.active
ws.append(["Plate Number"])

Process video

video_capture = cv2.VideoCapture("your_video.mp4")
while True:
ret, frame = video_capture.read()
if not ret:
break

# Perform object detection on the frame
blob = cv2.dnn.blobFromImage(frame, 0.00392, (416, 416), (0, 0, 0), True, crop=False)
net.setInput(blob)
outs = net.forward(output_layers)

# Extract bounding boxes and filter character detections
bboxes = []
for out in outs:
    for detection in out:
        scores = detection[5:]
        class_id = np.argmax(scores)
        confidence = scores[class_id]
        if confidence > 0.5:  # Confidence threshold
            center_x = int(detection[0] * frame.shape[1])
            center_y = int(detection[1] * frame.shape[0])
            w = int(detection[2] * frame.shape[1])
            h = int(detection[3] * frame.shape[0])
            x = int(center_x - w / 2)
            y = int(center_y - h / 2)
            if classes[class_id] in classes:
                bboxes.append((x, y, x + w, y + h))

# Group characters
groups = group_characters(bboxes)

# Extract plate numbers and store in Excel worksheet
for group in groups:
    plate_number = ''.join([classes[idx] for idx in group])
    ws.append([plate_number])

# Draw bounding boxes and display
for bbox in bboxes:
    cv2.rectangle(frame, (bbox[0], bbox[1]), (bbox[2], bbox[3]), (0, 255, 0), 2)
cv2.imshow("ANPR", frame)
if cv2.waitKey(1) & 0xFF == ord('q'):
    break

Save Excel workbook

wb.save("plate_numbers.xlsx")

video_capture.release()
cv2.destroyAllWindows()

[mintu07ruet]

glenn-jocher@ you are right! the url is too complex. I also used to ' ' but did not work, same error.

Here is my video link: https://mavsuta-my.sharepoint.com/:v:/r/personal/nrgomez_uta_edu/Documents/Mdmintu%20Video%20Experiment/Exp_2021630_12_55_2.avi?csf=1&web=1&e=kgbjLo/Exp_2021630_12_55_2.avi

if you want to test! Thanks for the help!

[glenn-jocher]

@mintu07ruet ah I see. In your case the URL is not publically available for download. If you click your link in a browser a download does not start, it is hidden behind a login screen, so naturally this video will not work. detect.py will work with any publically available image or video URL like https://ultralytics.com/assets/decelera_landscape.mov

[glenn-jocher]

@mintu07ruet I just pushed #5525 to improve URL inference. The following two examples will work well now, i.e. you can test them in our Colab notebook:
https://colab.research.google.com/github/ultralytics/yolov5/blob/master/tutorial.ipynb

!python detect.py --weights yolov5s.pt --source https://ultralytics.com/assets/zidane.jpg  # image URL
!python detect.py --weights yolov5s.pt --source https://ultralytics.com/assets/decelera_landscape.mov  # video URL

[mintu07ruet]

Thank you gleen-jocher@ for all the explanation. I will check to see if we can make our project video data public. But it is confidential in nature and big amount of video data like thousands of hour data that need to evaluate. We dumped all the video data in one drive which can share a particular person, may be not publicly. We prefer to run yolov5 in our local machine to save the output quickly. That's why we are not using kaggle and colab due to limitation of their GPU use.

If you have any updates on how to run the yolov5s.pt to run shared video from one drive directly in future. Please update me. Thank you so much for your time, help and suggestions! will back if I face further problems! Thanks!

[glenn-jocher]

@mintu07ruet we don't have the time to view your custom data, but in general you should run inference and training on local data for fastest results and lowest deployment risk.

[saktheeswaranswan]

Hi,

I am new to coding and posting first time on GitHub.

Well, I was able to successfully train the model on the custom dataset and I am using it to detect some signs in the videos.
I can see the results in the video as well as in the terminal.

Now I want my results to be saved in an excel file which should have a log of all the objects detected along with the time of the video.

Can someone please suggest how it can be done?

Here is all you want with the code its in my you tube channel bi have made an video on how to save yolo object detection with time stamp try to understand it the git link is given below the video description
https://youtu.be/1pw18Ml_SQ4

[saktheeswaranswan]

Here is the another code with some error correction and the code works good the if you do it on pandas it will work good and in Linux it gives permission to excel file but in windows ir shows permission error 13 which I have tried my best to rectify so check the links below first one shows about the code and the second one gives the permission error 13 in excel if it comes how to rectify it there is also another repo which uses pandas and csv with object detection tracking I'd with vehicle speed and time one of my subscriber told me that yolo v5 .pt pytorch weights also works on that which Is given in the video description so check it

https://youtu.be/360xRAnk1O0
Permission error 13 in excel which comes 8n windows give it acess and if it shows up move to ubantu Linux and run the code that's all I can do please support me on youtube by subscribing
https://youtu.be/fbsP4Az-LyQ

My final advice would be get administrator acess to Microsoft excel or use open office excel even if it shows error give full rights and run it error 13 permission denied for excel will not show up I guess this is my small contribution to our open source community

[saktheeswaranswan]

happy news problem solved but subscribe to random neural monk YouTube channel this works without showing perrmission 13 error so enjoy to log csv with time stamp class id also made with chat gpt took 6 hours

https://github.com/saktheeswaranswan/yolov5-csv-yolov5excel-crop-needs-yolov5-rasberrypi-GPIO

[Emmanuel-Muroboka]

Hello Mr @ glenn-jocher i have a question that am in need of your help with, is it possible to get the object detected information, say the names of a detected class, and display it on a different platform using the OPC-UA system? I have two computers one is running object detection which is our OPC client then the other computer is running the OPC server. I have created a node on the server which then the client subscribes to and can change the values of the node, so what I would want to show is the results am obtaining from the object detection to show on the server. How can I achieve this?