You are just mind-blowing what a brilliant and talented teacher before learning all this Machine learning deep learning reinforcement learning we need to learn basics Algorithms first then only we can understand so many problems and formulas need to solve in this learnings..Bow down
This is great! One question, YOLO split the picture into smaller one and look them only once. This seems better than that. Can you explain the difference? pros and cons? I'm also not very clear about how to run with different smaller window size.
What do you think about setting a threshold level below 0.5 in some specific tasks? Currently, I am training object detector for search and rescue drone, that will be looking for lost people. When I set the threshold level on lower level, for example 0.25, it works starts to look really better. I think it is because the main purpose, in this case of the detector, is to give a drone operator hint where a lost person could be, not to give him exact location. If the detector mark only half of the body of the lost person it still will be helpful for the operator. What do you think about it, could lowering of threshold level be useful and right in some specific tasks?
No, here we are training the algo and we need to improve the detection skill. Lowering the threshold will only result in a less effectively trained algo
Andrew, IoU for these awesome videos. :) Thanks!
I owe you, Andrew, you are an awesome teacher!
👏
You are just mind-blowing what a brilliant and talented teacher before learning all this Machine learning deep learning reinforcement learning we need to learn basics Algorithms first then only we can understand so many problems and formulas need to solve in this learnings..Bow down
concept delivered.....bullz eye in short span of time which is beautiful. Amazing stuff!!!
This is great! One question, YOLO split the picture into smaller one and look them only once. This seems better than that. Can you explain the difference? pros and cons? I'm also not very clear about how to run with different smaller window size.
Thank you very much for the lectures! Some of the lectures in C4W3 are out of order, though. Keep up the great work and sharing it with the world!!
The C4W4L05 is missing on the playlist and channel
Please upload C4W3L05
ua-cam.com/video/gKreZOUi-O0/v-deo.html
What do you think about setting a threshold level below 0.5 in some specific tasks? Currently, I am training object detector for search and rescue drone, that will be looking for lost people. When I set the threshold level on lower level, for example 0.25, it works starts to look really better. I think it is because the main purpose, in this case of the detector, is to give a drone operator hint where a lost person could be, not to give him exact location. If the detector mark only half of the body of the lost person it still will be helpful for the operator. What do you think about it, could lowering of threshold level be useful and right in some specific tasks?
No, here we are training the algo and we need to improve the detection skill. Lowering the threshold will only result in a less effectively trained algo
Great as always sir Andrew!
3:45 omg the joke :>>> Love you Andrew!
thats amazing and simple explanation! thanksss
This will only work on rectangles. How to calculate IoU of image segment of different shape?
output of the cnn is of the form(bx, by, bw, bw) which is a rectangle, so there will be no different shape.
I owe U, boy
l5 is missing
ground truth image means image in the training set right?
Yes It's the label of a training instance. Here the coordinates of the bounding box
Andrew is also fun. IoU I owe you.
nice expression
updated link to C4W3L05: ua-cam.com/video/gKreZOUi-O0/v-deo.html
Andrew's desk always looks so empty.