ClimbingMachine
commented
2 years ago I think Argoverse-API considers two inner lanes in the opposite direction (two-way traffic) as l_neighbor_ids in this case.
Open LaLaLailalai opened 3 years ago
ClimbingMachine
commented
2 years ago I think Argoverse-API considers two inner lanes in the opposite direction (two-way traffic) as l_neighbor_ids in this case.
I printed the value of graph["left_pairs"] and graph["right_pairs"] and found that in graph["left_pairs"] there are some duplicate node pairs, for example "[37,64]" and "[64,37]", but it is hard to understand why node "37" and "64" can be left neighbor of each other. But there are no repeated pairs in the graph["right_pairs"]. So why? If graph["left_pairs"] and the reverse of graph["right_pairs"] should be equal? For instance, graph["left_pairs"]:[[1,2],[3,4]], graph["right_pairs"]:[[2,1],[4,3]].
The printed value of graph["left_pairs"] and graph["right_pairs"] are as follows:
len graph[left_pairs] 67 len graph[right_pairs] 27 graph[left_pairs] tensor([[15, 91], [19, 86], [24, 97], [29, 30], [30, 29], [31, 43], [32, 68], [33, 60], [34, 39], [35, 51], [36, 66], [37, 64], [38, 69], [39, 34], [40, 67], [41, 95], [42, 51], [43, 31], [44, 29], [45, 78], [46, 63], [47, 69], [48, 58], [49, 67], [50, 58], [51, 35], [52, 15], [53, 47], [54, 37], [55, 56], [56, 55], [57, 96], [58, 48], [59, 75], [61, 48], [62, 82], [63, 72], [64, 37], [65, 49], [66, 98], [67, 49], [68, 32], [69, 47], [70, 64], [71, 30], [72, 63], [73, 74], [74, 82], [75, 59], [76, 77], [77, 76], [79, 85], [81, 80], [82, 74], [83, 85], [84, 91], [85, 83], [88, 83], [89, 35], [90, 72], [91, 15], [93, 92], [94, 98], [95, 41], [96, 57], [97, 24], [98, 66]], device='cuda:0') graph[right_pairs] tensor([[15, 52], [29, 44], [30, 71], [35, 89], [37, 54], [47, 53], [48, 61], [49, 65], [51, 42], [58, 50], [60, 33], [63, 46], [64, 70], [66, 36], [67, 40], [69, 38], [72, 90], [74, 73], [78, 45], [80, 81], [82, 62], [83, 88], [85, 79], [86, 19], [91, 84], [92, 93], [98, 94]], device='cuda:0')