How to fit non-squared input?

[shenxiaoang]

Hi, the original model takes 640x640 squared image as input, but here I really need non-squared input (e.g. 640x360) in my own project. I made some modifications to PriorBox class, but it seems that the trained model cannot work properly.

class PriorBox(object):
    def __init__(self, cfg, image_size=None, phase='train'):
        super(PriorBox, self).__init__()
        self.min_sizes = cfg['min_sizes']
        self.steps = cfg['steps']
        self.clip = cfg['clip']
        self.image_size = image_size
        self.feature_maps = [[ceil(self.image_size[0]/step), ceil(self.image_size[1]/step)] for step in self.steps]
        self.name = "s"

    def forward(self):
        anchors = []
        for k, f in enumerate(self.feature_maps):
            min_sizes = self.min_sizes[k]
            for i, j in product(range(int(f[0])), range(int(f[1]))):
                for min_size in min_sizes:
                    s_kx = min_size / (1.0*self.image_size[1])
                    s_ky = min_size / (1.0*self.image_size[1])
                    dense_cx = [x * self.steps[k] / self.image_size[0] for x in [i + 0.5]]
                    dense_cy = [y * self.steps[k] / self.image_size[1] for y in [j + 0.5]]
                    for cy, cx in product(dense_cy, dense_cx):
                        anchors += [cx, cy, s_kx, s_ky]
        # back to torch land
        output = torch.Tensor(anchors).view(-1, 4)
        if self.clip:
            output.clamp_(max=1, min=0)
        return output

So can you help me figure out the problem? Really appreciate it!