Excellent diagram with the multiple outputs, thanks for sharing. Just one observation, in the heatmap focal loss equation it is wrong it should be: L_k = -\frac{1}{N} \sum_{xyc}{\begin{cases} (1-\hat{Y}_{xyc})^{a}\;log(\hat{Y}_{xyc}) & \text{if } Y_{xyc} = 1\\ (1-Y_{xyc})^{\beta}(\hat{Y}_{xyc})^{\alpha}\;log(1-\hat{Y}_{xyc}) & \text{otherwise}\end{cases}}
📚 LINK TO BLOGPOST: learnopencv.com/centernet-anchor-free-object-detection-explained/ ▶ LINK TO OBJEST DETECTION PLAYLIST: ruclips.net/p/PLfYPZalDvZDK369GIdMp7Je5bPezs4hVQ ▶ LINK TO YOLO MASTERCLASS PLAYLIST: ruclips.net/p/PLfYPZalDvZDLALsG9o-cjwNelh-oW9Xc4
Hi, CenterNet is one of the foundation papers for anchorfree object detection. When it was released, it was compared with YOLOv3, Faster RCNN and RetinaNet.
@@LearnOpenCV Ahh ty for the reply. So how does it compare to newer yolo-versions ? Are there any noticable benefits over yolov7 or what's the best algorithm for object detection nowadays? Seems like things change daily since the ki boom
i have this issue when i work in YOLOV3 (tensorflow) G:\YOLOv3>python video_demo.py Traceback (most recent call last): File "G:\YOLOv3\video_demo.py", line 29, in return_tensors = utils.read_pb_return_tensors(graph, pb_file, return_elements) AttributeError: module 'core.utils' has no attribute 'read_pb_return_tensors'
Alone key points detection is not sufficient to measure size. However, if the object is at a fixed distance from the camera and there is a reference object size measurement at the same distance, we can calculate the object's size.
Yes, it is possible. When using OpenCV for annotations, you need to create a tuple of 80 colors (for COCO classes). For each class, you will assign a different tuple when drawing the bounding box around the object.
Excellent diagram with the multiple outputs, thanks for sharing. Just one observation, in the heatmap focal loss equation it is wrong it should be: L_k = -\frac{1}{N} \sum_{xyc}{\begin{cases}
(1-\hat{Y}_{xyc})^{a}\;log(\hat{Y}_{xyc}) & \text{if } Y_{xyc} = 1\\
(1-Y_{xyc})^{\beta}(\hat{Y}_{xyc})^{\alpha}\;log(1-\hat{Y}_{xyc}) & \text{otherwise}\end{cases}}
Thank you so much for correcting us. You have a sharp eye.
📚 LINK TO BLOGPOST: learnopencv.com/centernet-anchor-free-object-detection-explained/
▶ LINK TO OBJEST DETECTION PLAYLIST: ruclips.net/p/PLfYPZalDvZDK369GIdMp7Je5bPezs4hVQ
▶ LINK TO YOLO MASTERCLASS PLAYLIST: ruclips.net/p/PLfYPZalDvZDLALsG9o-cjwNelh-oW9Xc4
Question on 4:27 why the comparison with yolov3 instead of a more up to date model like v7 , v8 ?
Hi, CenterNet is one of the foundation papers for anchorfree object detection. When it was released, it was compared with YOLOv3, Faster RCNN and RetinaNet.
@@LearnOpenCV Ahh ty for the reply. So how does it compare to newer yolo-versions ? Are there any noticable benefits over yolov7 or what's the best algorithm for object detection nowadays? Seems like things change daily since the ki boom
i have this issue when i work in YOLOV3
(tensorflow) G:\YOLOv3>python video_demo.py
Traceback (most recent call last):
File "G:\YOLOv3\video_demo.py", line 29, in
return_tensors = utils.read_pb_return_tensors(graph, pb_file, return_elements)
AttributeError: module 'core.utils' has no attribute 'read_pb_return_tensors'
Excellent Video. Thank you
You are welcome!
Excellent explanation!
Glad you think so!
Hi Sir, excellent explanation, can the size of an object be measured using the key point detection
Alone key points detection is not sufficient to measure size. However, if the object is at a fixed distance from the camera and there is a reference object size measurement at the same distance, we can calculate the object's size.
@@prakashchandra90 Thank you, Prakash. I am doing a project on calculating an object's height, give me some insights if known
It is possible to change color each classe of bounding boxes ??
Yes, it is possible. When using OpenCV for annotations, you need to create a tuple of 80 colors (for COCO classes). For each class, you will assign a different tuple when drawing the bounding box around the object.
@@sovitrath4735 do you havé discord i want to send you my Own code about it. If you can help me to resolve it
I love yolo
We do too! That's why we have a full-fledged playlist on YOLO on our channel. Do check it out!