SIFT，SuperPoint在图像特征提取上的对比实验

SIFT，SuperPoint都具有提取图片特征点，并且输出特征描述子的特性，本篇文章从特征点的提取数量，特征点的正确匹配数量来探索一下二者的优劣。

视角变化较大的情况下

原图1	原图2	SuperPoint特征点数	SIFT提取到的特征点数

SuperPoint特征点匹配情况	SIFT特征点匹配情况

• SuperPoint提取到的特征点数量要少一些，可以理解，我想原因大概是SuperPoint训练使用的是合成数据集，含有很多形状，并且只标出了线段的一些拐点，而sift对图像的像素值变化敏感。

SuperPoint下MagicPoint训练集样例

• 在特征点匹配上，感觉不出有什么明显的差异，但是很明显，SuperPoint的鲁棒性更高一些，sift匹配有很多的错点，比如SIFT第三幅图中的牛奶盒子，由于物体没有上下的起伏，可以认为连线中的斜线都是错匹配。

在形状较为复杂的情况下

正如上文所说，SuperPoint对形状较多的图片敏感。

原图1	原图2	SuperPoint特征点数	SIFT提取到的特征点数

SuperPoint特征点匹配情况	SIFT特征点匹配情况

这里也可以看出，SuperPoint的匹配精度要优于SIFT，不会出现很多杂乱的线。

在3D建模object的表现

我尝试了之前自己的建模，并且观察了不同旋转角度对特征匹配的影响

主视角		小角度	视角缩放	旋转角度过大时
	SuperPoint
	SIFT
		此时二者差距不大	这个虽然sift匹配的点更多，但是基本上都是错的。	这个虽然sift匹配的点更多，但是基本上都是错的。

SIFT特征点检测情况对比：

SuperPoint特征点检测情况对比：

同样值得注意的是，第一张图的窗子的点，SuperPoint并没有检测出来。

总结

• 在捕捉特征点的时候，SuperPoint对形状的特征点敏感，SIFT对像素的变化敏感
• 在进行特征点匹配的时候，SuperPoint的特征描述子鲁棒性更好一些
• 视角变化较大的情况下，二者的表现都差强人意

代码

SIFT.py：

from __future__ import print_function
import cv2 as cv
import numpy as np
import argparse

pic1 = "./1.ppm"
pic2 = "./6.ppm"

parser = argparse.ArgumentParser(description='Code for Feature Matching with FLANN tutorial.')
parser.add_argument('--input1', help='Path to input image 1.', default=pic1)
parser.add_argument('--input2', help='Path to input image 2.', default=pic2)
args = parser.parse_args()
img_object = cv.imread(pic1)
img_scene = cv.imread(pic2)
if img_object is None or img_scene is None:
    print('Could not open or find the images!')
    exit(0)

#-- Step 1: Detect the keypoints using SURF Detector, compute the descriptors
minHessian = 600
detector = cv.xfeatures2d_SURF.create(hessianThreshold=minHessian)
keypoints_obj, descriptors_obj = detector.detectAndCompute(img_object, None)
keypoints_scene, descriptors_scene = detector.detectAndCompute(img_scene, None)

#-- Step 2: Matching descriptor vectors with a FLANN based matcher
# Since SURF is a floating-point descriptor NORM_L2 is used
matcher = cv.DescriptorMatcher_create(cv.DescriptorMatcher_FLANNBASED)
knn_matches = matcher.knnMatch(descriptors_obj, descriptors_scene, 2)

#-- Filter matches using the Lowe's ratio test
ratio_thresh = 0.75
good_matches = []
for m,n in knn_matches:
    if m.distance < ratio_thresh * n.distance:
        good_matches.append(m)

print("The number of keypoints in image1 is", len(keypoints_obj))
print("The number of keypoints in image2 is", len(keypoints_scene))
#-- Draw matches
img_matches = np.empty((max(img_object.shape[0], img_scene.shape[0]), img_object.shape[1]+img_scene.shape[1], 3), dtype=np.uint8)
cv.drawMatches(img_object, keypoints_obj, img_scene, keypoints_scene, good_matches, img_matches, flags=cv.DrawMatchesFlags_NOT_DRAW_SINGLE_POINTS)

cv.namedWindow("Good Matches of SIFT", 0)
cv.resizeWindow("Good Matches of SIFT", 1024, 1024)
cv.imshow('Good Matches of SIFT', img_matches)
cv.waitKey()

使用sift.py时，只需要修改第6,7行的图片路径即可。

SuperPoint

import numpy as np
import os
import cv2
import torch

# Jet colormap for visualization.
myjet = np.array([[0., 0., 0.5],
                  [0., 0., 0.99910873],
                  [0., 0.37843137, 1.],
                  [0., 0.83333333, 1.],
                  [0.30044276, 1., 0.66729918],
                  [0.66729918, 1., 0.30044276],
                  [1., 0.90123457, 0.],
                  [1., 0.48002905, 0.],
                  [0.99910873, 0.07334786, 0.],
                  [0.5, 0., 0.]])

class SuperPointNet(torch.nn.Module):
    """ Pytorch definition of SuperPoint Network. """

    def __init__(self):
        super(SuperPointNet, self).__init__()
        self.relu = torch.nn.ReLU(inplace=True)
        self.pool = torch.nn.MaxPool2d(kernel_size=2, stride=2)
        c1, c2, c3, c4, c5, d1 = 64, 64, 128, 128, 256, 256
        # Shared Encoder.
        self.conv1a = torch.nn.Conv2d(1, c1, kernel_size=3, stride=1, padding=1)
        self.conv1b = torch.nn.Conv2d(c1, c1, kernel_size=3, stride=1, padding=1)
        self.conv2a = torch.nn.Conv2d(c1, c2, kernel_size=3, stride=1, padding=1)
        self.conv2b = torch.nn.Conv2d(c2, c2, kernel_size=3, stride=1, padding=1)
        self.conv3a = torch.nn.Conv2d(c2, c3, kernel_size=3, stride=1, padding=1)
        self.conv3b = torch.nn.Conv2d(c3, c3, kernel_size=3, stride=1, padding=1)
        self.conv4a = torch.nn.Conv2d(c3, c4, kernel_size=3, stride=1, padding=1)
        self.conv4b = torch.nn.Conv2d(c4, c4, kernel_size=3, stride=1, padding=1)
        # Detector Head.
        self.convPa = torch.nn.Conv2d(c4, c5, kernel_size=3, stride=1, padding=1)
        self.convPb = torch.nn.Conv2d(c5, 65, kernel_size=1, stride=1, padding=0)
        # Descriptor Head.
        self.convDa = torch.nn.Conv2d(c4, c5, kernel_size=3, stride=1, padding=1)
        self.convDb = torch.nn.Conv2d(c5, d1, kernel_size=1, stride=1, padding=0)

    def forward(self, x):
        """ Forward pass that jointly computes unprocessed point and descriptor
        tensors.
        Input
          x: Image pytorch tensor shaped N x 1 x H x W.
        Output
          semi: Output point pytorch tensor shaped N x 65 x H/8 x W/8.
          desc: Output descriptor pytorch tensor shaped N x 256 x H/8 x W/8.
        """
        # Shared Encoder.
        x = self.relu(self.conv1a(x))
        x = self.relu(self.conv1b(x))
        x = self.pool(x)
        x = self.relu(self.conv2a(x))
        x = self.relu(self.conv2b(x))
        x = self.pool(x)
        x = self.relu(self.conv3a(x))
        x = self.relu(self.conv3b(x))
        x = self.pool(x)
        x = self.relu(self.conv4a(x))
        x = self.relu(self.conv4b(x))
        # Detector Head.
        cPa = self.relu(self.convPa(x))
        semi = self.convPb(cPa)
        # Descriptor Head.
        cDa = self.relu(self.convDa(x))
        desc = self.convDb(cDa)
        dn = torch.norm(desc, p=2, dim=1)  # Compute the norm.
        desc = desc.div(torch.unsqueeze(dn, 1))  # Divide by norm to normalize.
        return semi, desc

class SuperPointFrontend(object):
    """ Wrapper around pytorch net to help with pre and post image processing. """

    def __init__(self, weights_path, nms_dist, conf_thresh, nn_thresh,
                 cuda=False):
        self.name = 'SuperPoint'
        self.cuda = cuda
        self.nms_dist = nms_dist
        self.conf_thresh = conf_thresh
        self.nn_thresh = nn_thresh  # L2 descriptor distance for good match.
        self.cell = 8  # Size of each output cell. Keep this fixed.
        self.border_remove = 4  # Remove points this close to the border.

        # Load the network in inference mode.
        self.net = SuperPointNet()
        if cuda:
            # Train on GPU, deploy on GPU.
            self.net.load_state_dict(torch.load(weights_path))
            self.net = self.net.cuda()
        else:
            # Train on GPU, deploy on CPU.
            self.net.load_state_dict(torch.load(weights_path,
                                                map_location=lambda storage, loc: storage))
        self.net.eval()

    def nms_fast(self, in_corners, H, W, dist_thresh):
        """
        Run a faster approximate Non-Max-Suppression on numpy corners shaped:
          3xN [x_i,y_i,conf_i]^T

        Algo summary: Create a grid sized HxW. Assign each corner location a 1, rest
        are zeros. Iterate through all the 1's and convert them either to -1 or 0.
        Suppress points by setting nearby values to 0.

        Grid Value Legend:
        -1 : Kept.
         0 : Empty or suppressed.
         1 : To be processed (converted to either kept or supressed).

        NOTE: The NMS first rounds points to integers, so NMS distance might not
        be exactly dist_thresh. It also assumes points are within image boundaries.

        Inputs
          in_corners - 3xN numpy array with corners [x_i, y_i, confidence_i]^T.
          H - Image height.
          W - Image width.
          dist_thresh - Distance to suppress, measured as an infinty norm distance.
        Returns
          nmsed_corners - 3xN numpy matrix with surviving corners.
          nmsed_inds - N length numpy vector with surviving corner indices.
        """
        grid = np.zeros((H, W)).astype(int)  # Track NMS data.
        inds = np.zeros((H, W)).astype(int)  # Store indices of points.
        # Sort by confidence and round to nearest int.
        inds1 = np.argsort(-in_corners[2, :])
        corners = in_corners[:, inds1]
        rcorners = corners[:2, :].round().astype(int)  # Rounded corners.
        # Check for edge case of 0 or 1 corners.
        if rcorners.shape[1] == 0:
            return np.zeros((3, 0)).astype(int), np.zeros(0).astype(int)
        if rcorners.shape[1] == 1:
            out = np.vstack((rcorners, in_corners[2])).reshape(3, 1)
            return out, np.zeros((1)).astype(int)
        # Initialize the grid.
        for i, rc in enumerate(rcorners.T):
            grid[rcorners[1, i], rcorners[0, i]] = 1
            inds[rcorners[1, i], rcorners[0, i]] = i
        # Pad the border of the grid, so that we can NMS points near the border.
        pad = dist_thresh
        grid = np.pad(grid, ((pad, pad), (pad, pad)), mode='constant')
        # Iterate through points, highest to lowest conf, suppress neighborhood.
        count = 0
        for i, rc in enumerate(rcorners.T):
            # Account for top and left padding.
            pt = (rc[0] + pad, rc[1] + pad)
            if grid[pt[1], pt[0]] == 1:  # If not yet suppressed.
                grid[pt[1] - pad:pt[1] + pad + 1, pt[0] - pad:pt[0] + pad + 1] = 0
                grid[pt[1], pt[0]] = -1
                count += 1
        # Get all surviving -1's and return sorted array of remaining corners.
        keepy, keepx = np.where(grid == -1)
        keepy, keepx = keepy - pad, keepx - pad
        inds_keep = inds[keepy, keepx]
        out = corners[:, inds_keep]
        values = out[-1, :]
        inds2 = np.argsort(-values)
        out = out[:, inds2]
        out_inds = inds1[inds_keep[inds2]]
        return out, out_inds

    def run(self, img):
        """ Process a numpy image to extract points and descriptors.
        Input
          img - HxW numpy float32 input image in range [0,1].
        Output
          corners - 3xN numpy array with corners [x_i, y_i, confidence_i]^T.
          desc - 256xN numpy array of corresponding unit normalized descriptors.
          heatmap - HxW numpy heatmap in range [0,1] of point confidences.
          """
        assert img.ndim == 2, 'Image must be grayscale.'
        assert img.dtype == np.float32, 'Image must be float32.'
        H, W = img.shape[0], img.shape[1]
        inp = img.copy()
        inp = (inp.reshape(1, H, W))
        inp = torch.from_numpy(inp)
        inp = torch.autograd.Variable(inp).view(1, 1, H, W)
        if self.cuda:
            inp = inp.cuda()
        # Forward pass of network.
        outs = self.net.forward(inp)
        semi, coarse_desc = outs[0], outs[1]
        # Convert pytorch -> numpy.
        semi = semi.data.cpu().numpy().squeeze()
        # --- Process points.
        # C = np.max(semi)
        # dense = np.exp(semi - C)  # Softmax.
        # dense = dense / (np.sum(dense))  # Should sum to 1.
        dense = np.exp(semi)  # Softmax.
        dense = dense / (np.sum(dense, axis=0) + .00001)  # Should sum to 1.
        # Remove dustbin.
        nodust = dense[:-1, :, :]
        # Reshape to get full resolution heatmap.
        Hc = int(H / self.cell)
        Wc = int(W / self.cell)
        nodust = nodust.transpose(1, 2, 0)
        heatmap = np.reshape(nodust, [Hc, Wc, self.cell, self.cell])
        heatmap = np.transpose(heatmap, [0, 2, 1, 3])
        heatmap = np.reshape(heatmap, [Hc * self.cell, Wc * self.cell])
        xs, ys = np.where(heatmap >= self.conf_thresh)  # Confidence threshold.
        if len(xs) == 0:
            return np.zeros((3, 0)), None, None
        pts = np.zeros((3, len(xs)))  # Populate point data sized 3xN.
        pts[0, :] = ys
        pts[1, :] = xs
        pts[2, :] = heatmap[xs, ys]
        pts, _ = self.nms_fast(pts, H, W, dist_thresh=self.nms_dist)  # Apply NMS.
        inds = np.argsort(pts[2, :])
        pts = pts[:, inds[::-1]]  # Sort by confidence.
        # Remove points along border.
        bord = self.border_remove
        toremoveW = np.logical_or(pts[0, :] < bord, pts[0, :] >= (W - bord))
        toremoveH = np.logical_or(pts[1, :] < bord, pts[1, :] >= (H - bord))
        toremove = np.logical_or(toremoveW, toremoveH)
        pts = pts[:, ~toremove]
        # --- Process descriptor.
        D = coarse_desc.shape[1]
        if pts.shape[1] == 0:
            desc = np.zeros((D, 0))
        else:
            # Interpolate into descriptor map using 2D point locations.
            samp_pts = torch.from_numpy(pts[:2, :].copy())
            samp_pts[0, :] = (samp_pts[0, :] / (float(W) / 2.)) - 1.
            samp_pts[1, :] = (samp_pts[1, :] / (float(H) / 2.)) - 1.
            samp_pts = samp_pts.transpose(0, 1).contiguous()
            samp_pts = samp_pts.view(1, 1, -1, 2)
            samp_pts = samp_pts.float()
            if self.cuda:
                samp_pts = samp_pts.cuda()
            desc = torch.nn.functional.grid_sample(coarse_desc, samp_pts)
            desc = desc.data.cpu().numpy().reshape(D, -1)
            desc /= np.linalg.norm(desc, axis=0)[np.newaxis, :]
        return pts, desc, heatmap

if __name__ == '__main__':

    print('==> Loading pre-trained network.')
    # This class runs the SuperPoint network and processes its outputs.
    fe = SuperPointFrontend(weights_path="superpoint_v1.pth",
                            nms_dist=4,
                            conf_thresh=0.015,
                            nn_thresh=0.7,
                            cuda=True)
    print('==> Successfully loaded pre-trained network.')

    pic1 = "./1.ppm"
    pic2 = "./6.ppm"

    image1_origin = cv2.imread(pic1)
    image2_origin = cv2.imread(pic2)

    image1 = cv2.imread(pic1, cv2.IMREAD_GRAYSCALE).astype(np.float32)
    image2 = cv2.imread(pic2, cv2.IMREAD_GRAYSCALE).astype(np.float32)
    image1 = image1 / 255.
    image2 = image2 / 255.

    if image1 is None or image2 is None:
        print('Could not open or find the images!')
        exit(0)

    # -- Step 1: Detect the keypoints using SURF Detector, compute the descriptors

    keypoints_obj, descriptors_obj, h1 = fe.run(image1)
    keypoints_scene, descriptors_scene, h2 = fe.run(image2)

    ## to transfer array ==> KeyPoints
    keypoints_obj = [cv2.KeyPoint(keypoints_obj[0][i], keypoints_obj[1][i], 1)
                for i in range(keypoints_obj.shape[1])]
    keypoints_scene = [cv2.KeyPoint(keypoints_scene[0][i], keypoints_scene[1][i], 1)
                     for i in range(keypoints_scene.shape[1])]
    print("The number of keypoints in image1 is", len(keypoints_obj))
    print("The number of keypoints in image2 is", len(keypoints_scene))

    # -- Step 2: Matching descriptor vectors with a FLANN based matcher
    # Since SURF is a floating-point descriptor NORM_L2 is used
    matcher = cv2.DescriptorMatcher_create(cv2.DescriptorMatcher_FLANNBASED)
    knn_matches = matcher.knnMatch(descriptors_obj.T, descriptors_scene.T, 2)

    # -- Filter matches using the Lowe's ratio test
    ratio_thresh = 0.75
    good_matches = []
    for m, n in knn_matches:
        if m.distance < ratio_thresh * n.distance:
            good_matches.append(m)

  # -- Draw matches
    img_matches = np.empty((max(image1_origin.shape[0], image2_origin.shape[0]), image1_origin.shape[1] + image2_origin.shape[1], 3),
                           dtype=np.uint8)
    cv2.drawMatches(image1_origin, keypoints_obj, image2_origin, keypoints_scene, good_matches, img_matches,
                    flags=cv2.DrawMatchesFlags_NOT_DRAW_SINGLE_POINTS)

    cv2.namedWindow("Good Matches of SuperPoint", 0)
    cv2.resizeWindow("Good Matches of SuperPoint", 1024, 1024)
    cv2.imshow('Good Matches of SuperPoint', img_matches)
    cv2.waitKey()

superpoint.py是基于官方给出的代码修改得到，使用步骤如下：

1. 去官网下载模型的预训练文件，https://github.com/magicleap/SuperPointPretrainedNetwork

1. 将预训练文件与文中给出的superpoint代码放在统一目录下

   

2. 修改254,255行的图片路径运行即可