First of all, thank you for your question. The full name of TCN is Temporal Convolutional Network. The TCN model is based on the CNN model, and the following improvements have been made:
We finally chose to use the Temporal Convolution Network to train the model we needed. We use TCN for several reasons. First of all, the videos are sequential, and the exercise is coherent. Unlike pictures, which can be learned one by one, TCN can predict the output at time t is only convolved with the elements that occurred before t, without being affected by the subsequent data impact. Secondly, we adjusted the coefficient of dilation is 2, increase the size of the receptive field, reduce the position error of the characters in the video and the key points captured by the computer, and reduce the usage of RAM.
I hope my answer let you know the difference between CNN and TCN and why we chose to train our model with TCN.
An exciting and practical project to motivate people in sports! Plus, the UI and grading function is lit!!! Our group also proposed a similar project, named Power Fitness. Our system focused on simple poses in Strength, Bodyweight and Yoga. Would you mind taking a look at our poster? Appreciate!
Indeed. In fact, we are finding a way to host our website on the Cloud server. In the meantime, you may come to our Zoom Breakout room to see the real-time demo~
May I ask what is the difference between CNN and TCN?
First of all, thank you for your question. The full name of TCN is Temporal Convolutional Network. The TCN model is based on the CNN model, and the following improvements have been made:
Applicable sequence model: Causal Convolution
Memory history: Dilated Convolution, Residual block
We finally chose to use the Temporal Convolution Network to train the model we needed. We use TCN for several reasons. First of all, the videos are sequential, and the exercise is coherent. Unlike pictures, which can be learned one by one, TCN can predict the output at time t is only convolved with the elements that occurred before t, without being affected by the subsequent data impact. Secondly, we adjusted the coefficient of dilation is 2, increase the size of the receptive field, reduce the position error of the characters in the video and the key points captured by the computer, and reduce the usage of RAM.
I hope my answer let you know the difference between CNN and TCN and why we chose to train our model with TCN.
Very detailed explanation! Thank you.
An exciting and practical project to motivate people in sports! Plus, the UI and grading function is lit!!! Our group also proposed a similar project, named Power Fitness. Our system focused on simple poses in Strength, Bodyweight and Yoga. Would you mind taking a look at our poster? Appreciate!
Portal –> https://industryday.ece.ust.hk/2022/?p=3446
Thank you. Your project is also fascinating! May I have a chance to use your system?
Indeed. In fact, we are finding a way to host our website on the Cloud server. In the meantime, you may come to our Zoom Breakout room to see the real-time demo~
Great, I appreciate your invitation！
very good idea, love from HK
Thank you Chi Wan. God bless you always!!