Non-local_pytorch icon indicating copy to clipboard operation
Non-local_pytorch copied to clipboard

Published 20 hours ago • verified publisher icon AlexHex7

使用点积的热图可视化

Open 55renchen opened this issue 4 years ago • 1 comments

您好，我用点积版本的non_local对f进行softmax之后， f = torch.matmul(theta_x, phi_x) N = f.size(-1) f_div_C = f / N sft = nn.Softmax(dim=2) f_div_C = sft(f_div_C) 在可视化时还需要对NL_MAP进行normalize吗？如果需要，应该如何normalize呢？ 现在我的热图结果只有一个box，看不到正常热图颜色的变化，这是什么原因呢？

55renchen avatar Jun 15 '21 02:06 55renchen

Hi @55renchen , NL_MAP需要归一化到[0, 1]. uint8类型范围是[0, 255]，超出该范围则溢出

https://github.com/AlexHex7/Non-local_pytorch/blob/1fdae52352c90ef53b547129106103a20c8a36fd/nl_map_vis/nl_map_vis.py#L35

AlexHex7 avatar Jun 28 '21 02:06 AlexHex7