torch.size question in NavPedPreNet()

[SpikeJishuo]

DDRL4NAV/USTC_lab/nn/nav_encoder.py

Line 72 in 8d52815

encoded_image = self._encode_image(torch.cat([state[0], state[2]], axis=1))

Could you plz tell me what does state[2] stand for in pedstrians exsisting scenario?
I got torch.size([1, 96, 48]) of torch.cat([state[0], state[2]], axis=1))) instead of ([64, 4, 3, 3]) .

[SnallQiu]

DDRL4NAV/USTC_lab/nn/nav_encoder.py

Line 72 in 8d52815

encoded_image = self._encode_image(torch.cat([state[0], state[2]], axis=1))

Could you plz tell me what does state[2] stand for in pedstrians exsisting scenario? I got torch.size([1, 96, 48]) of torch.cat([state[0], state[2]], axis=1))) instead of ([64, 4, 3, 3]) .

the index of state here is defined in https://github.com/DRL-Navigation/img_env/blob/master/envs/wrapper/filter_states.py

• state[0]: means sensor_maps, [batch, 1, 48, 48], egocentric sensor map.
• state[2]: ped_maps, [batch, 3, 48, 48] , egocentric 3-channel pedestrians map, which means x，y velocity and position of the nearby pedestrians.

For more details of the state, you can see readme of https://github.com/DRL-Navigation/img_env.