Closed Deante-dx closed 4 months ago
Hello. The lines of the code you were mentioning were for data augmentation. Currently the dataset contains the data for the mocap markers with respect to 4 cameras. We added some imaginary cameras and converted our data to their coordinate space. It helped in order to improve the model accuracy.
Hello. The lines of the code you were mentioning were for data augmentation. Currently the dataset contains the data for the mocap markers with respect to 4 cameras. We added some imaginary cameras and converted our data to their coordinate space. It helped in order to improve the model accuracy.
Thanks for your response. As shown in the code below, augment with respect to a random camera does not have a new entry_key to save it, so it will overwrite the previous data. Why is it done this way?
`added_sact= 400
for aug in range(len(self._augment_cameras_ang)):
added_sact += 100
ang = self._augment_cameras_ang[aug]
x_cam = self._augment_cameras_x[aug]
y_cam = self._augment_cameras_y[aug]
## augment with respect to a random camera
new_seq, new_seq_traj = self.new_camera_data(s_id,correct_action,sact,added_sact,ang,x_cam,y_cam,action_sequence_main)
self._data[entry_key] = new_seq
self._traj[entry_key] = new_seq_traj`
@Deante-dx Sorry for the late reply. It doesn't overwrite them. It adds them to the total settings by another line of the code.
Hi there,
I've come across various camera-related operations. I would like to understand the significance of these operations and how they relate to the overall data processing pipeline. Could someone kindly provide a detailed explanation of the camera operations and their relevance in data processing?
codes in H36MDataset_v3
Thank you in advance for your assistance and guidance!
Best regards,