#1 Added weight conversion to Pytorch

convert_weights.py

@@ -0,0 +1,26 @@
+import copy
+import torch
+import numpy as np
+import megengine as mge
+
+from nets import Model
+
+# Read Megengine parameters
+pretrained_dict = mge.load("models/crestereo_eth3d.mge")
+
+model = Model(max_disp=256, mixed_precision=False, test_mode=True)
+model.eval()
+
+state_dict = model.state_dict()
+for key, value in pretrained_dict['state_dict'].items():
+
+	print(f"Converting {key}")
+	# Fix shape mismatch
+	if value.shape[0] == 1:
+		value = np.squeeze(value)
+
+	state_dict[key] = torch.tensor(value)
+
+output_path = "models/crestereo_eth3d.pth"
+torch.save(state_dict, output_path)
+print(f"\nModel saved to: {output_path}")

test_model.py

@@ -1,13 +1,63 @@
 import torch
+import torch.nn.functional as F
+import numpy as np
+import cv2
+from imread_from_url import imread_from_url
 
 from nets import Model
 
+#Ref: https://github.com/megvii-research/CREStereo/blob/master/test.py
+def inference(left, right, model, n_iter=20):
+	
+	print("Model Forwarding...")
+	imgL = left.transpose(2, 0, 1)
+	imgR = right.transpose(2, 0, 1)
+	imgL = np.ascontiguousarray(imgL[None, :, :, :])
+	imgR = np.ascontiguousarray(imgR[None, :, :, :])
+
+	imgL = torch.tensor(imgL.astype("float32"))
+	imgR = torch.tensor(imgR.astype("float32"))
+
+	imgL_dw2 = F.interpolate(
+		imgL,
+		size=(imgL.shape[2] // 2, imgL.shape[3] // 2),
+		mode="bilinear",
+		align_corners=True,
+	)
+	imgR_dw2 = F.interpolate(
+		imgR,
+		size=(imgL.shape[2] // 2, imgL.shape[3] // 2),
+		mode="bilinear",
+		align_corners=True,
+	)
+	# print(imgR_dw2.shape)
+
+	pred_flow_dw2 = model(imgL_dw2, imgR_dw2, iters=n_iter, flow_init=None)
+
+	pred_flow = model(imgL, imgR, iters=n_iter, flow_init=pred_flow_dw2)
+	pred_disp = torch.squeeze(pred_flow[:, 0, :, :]).detach().numpy()
+
+	return pred_disp
+
+if __name__ == '__main__':
+
+	left_img = imread_from_url("https://vision.middlebury.edu/stereo/data/scenes2003/newdata/cones/im2.png")
+	right_img = imread_from_url("https://vision.middlebury.edu/stereo/data/scenes2003/newdata/cones/im6.png")
+
+	model_path = "models/crestereo_eth3d.pth"
+
 	model = Model(max_disp=256, mixed_precision=False, test_mode=True)
+	model.load_state_dict(torch.load(model_path), strict=True)
 	model.eval()
 
-t1 = torch.rand(1, 3, 480, 640)
+	disp = inference(left_img, right_img, model, n_iter=20)
-t2 = torch.rand(1, 3, 480, 640)
+	disp_vis = (disp - disp.min()) / (disp.max() - disp.min()) * 255.0
+	disp_vis = disp_vis.astype("uint8")
+	disp_vis = cv2.applyColorMap(disp_vis, cv2.COLORMAP_INFERNO)
+
+	cv2.imshow("output", disp_vis)
+	cv2.imwrite("output.jpg", disp_vis)
+	cv2.waitKey(0)
+
 
-output = model(t1,t2)
-print(output.shape)