How are npy files in the directory "link_npy" generated?
Closed this issue · 1 comments
nek0peko commented
I printed npy[0] out and it shows:
{
'name': '101_alfonso',
'bbox': tensor([[3.2000e+01, 0.0000e+00, 4.6000e-01, 7.3333e-02, 5.4000e-01],
[1.8000e+01, 9.6667e-02, 4.9667e-01, 1.5000e-01, 5.2333e-01],
[1.0000e+00, 1.9667e-01, 4.0000e-01, 2.3000e-01, 5.8667e-01],
[6.7000e+01, 2.5333e-01, 4.5667e-01, 3.5667e-01, 5.2000e-01],
[8.2000e+01, 2.5000e-01, 5.4333e-01, 2.8333e-01, 5.8333e-01],
[1.8000e+01, 2.9667e-01, 5.6333e-01, 3.2667e-01, 5.8333e-01],
[9.0000e+00, 3.4000e-01, 5.5000e-01, 3.8000e-01, 5.8667e-01],
[8.7000e+01, 2.8000e-01, 4.1333e-01, 3.1333e-01, 4.4667e-01],
[6.9000e+01, 3.9333e-01, 4.5667e-01, 4.4000e-01, 5.1333e-01],
[8.2000e+01, 4.3333e-01, 5.0667e-01, 4.7000e-01, 5.4333e-01],
[5.0000e+00, 4.8000e-01, 4.9000e-01, 5.2667e-01, 4.9667e-01],
[1.0000e+00, 5.7333e-01, 4.0667e-01, 5.9667e-01, 5.5000e-01],
[8.7000e+01, 6.1000e-01, 4.6333e-01, 6.5333e-01, 5.1333e-01],
[5.0000e+00, 6.7333e-01, 4.8667e-01, 7.0667e-01, 4.9333e-01],
[1.0000e+01, 7.2000e-01, 4.5667e-01, 7.6667e-01, 5.0667e-01],
[2.0000e+00, 7.5000e-01, 4.0333e-01, 7.8667e-01, 5.5000e-01],
[6.0000e+01, 8.0333e-01, 4.5667e-01, 8.5333e-01, 5.1333e-01],
[2.0000e+00, 8.1000e-01, 3.8667e-01, 8.7000e-01, 6.0667e-01],
[9.1000e+01, 9.0333e-01, 4.6000e-01, 9.5000e-01, 5.2333e-01],
[1.0000e+01, 9.5667e-01, 4.0333e-01, 1.0000e+00, 4.5000e-01]]),
'edge_type': tensor([[0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[2, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 2, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 2, 0, 3, 0, 0, 4, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 4, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 2, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 2, 0, 0, 0, 0, 0, 3, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 4, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 2, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0, 1, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0, 1, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0, 1, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0, 1, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0, 1, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0, 1, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0, 1, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0, 4],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 3, 0]])
}
If I want to change to my own dataset, and my annotations are like:
[
{
"image": "xxxx.jpg",
"annotations": [
{
"label": "a",
"coordinates": {
"x": 78,
"y": 83.5,
"width": 29.5,
"height": 161
}
},
{
"label": "b",
"coordinates": {
"x": 120,
"y": 68.5,
"width": 31.5,
"height": 137
}
},
{
"label": "c",
"coordinates": {
"x": 57.5,
"y": 143,
"width": 33.5,
"height": 44.5
}
},
......
],
"edges": ["'a', 'b', 'right'", "'b', 'c', 'sub'"]
},
......
]
How to preprocess my dataset to apply your code?
I'd appreciate it if you can help me with this.
nek0peko commented
I changed the way to print it and understood the data structure now.
Types of symbols and relations are in vocab.json.
(dict_keys(['name', 'bbox', 'edge_type']),
{'name': '2_em_3',
'bbox': tensor([[71.0000, 0.0000, 0.1433, 0.2067, 0.8533],
[33.0000, 0.3267, 0.2133, 0.7567, 0.6367],
[72.0000, 0.9167, 0.1400, 1.0000, 0.8200]]),
'edge_type': tensor([[0, 1, 0],
[2, 0, 1],
[0, 2, 0]])})
