We have a trained deep learning model which takes initial pose ( i.e. coordinates of every joint in body) and some trajectories ( let's say a trajectory for right knee join and left shoulder joint) as an input to that model and that model will try to predict the poses for further time-frames according to the input data.
Now we want to create a web based application (with unity 3d or python Dash with Flask service), in which for eg there will three buttons pose1 pose2 pose3. When you click on them an initial pose will appear on the screen. There will be 17 joints in that pose because our dataset has 17 joints in the body. Now we want user to select any joint (let's say using a drop down menu ) and it should allow it to draw a trajectory starting from that selected joint. This function should be available for rest of the joints as well. After this, there will another button, let's say "animate" button. When user presses it, the data of initial pose and trajectory should go to our deep learning model using a Flask service or any other technology. The model will predict the next poses and send the data back for animation. After this user should see the animation on the screen.
Note :- Dataset used in our deep learning project is h3.6m dataset.