add: changes

.gitattributes

@@ -26,3 +26,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zstandard filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
 monoscene_kitti.ckpt filter=lfs diff=lfs merge=lfs -text
+monoscene_nyu.ckpt filter=lfs diff=lfs merge=lfs -text

app.py

@@ -22,10 +22,41 @@ model = MonoScene.load_from_checkpoint(
     full_scene_size=(60, 36, 60),
 )
 
-img_W, img_H = 640, 480
 
+def get_projections(img_W, img_H):
+    scale_3ds = [1, 2]
+    data = {}
+    for scale_3d in scale_3ds:
+        scene_size = (4.8, 4.8, 2.88)
+        vox_origin = np.array([-1.54591799,  0.8907361 , -0.05 ])
+        voxel_size = 0.08
+
+        
+        cam_k = np.array([[518.8579, 0, 320], [0, 518.8579, 240], [0, 0, 1]])
+        cam_pose = np.asarray([[ 9.6699458e-01,  4.2662762e-02,  2.5120059e-01,  0.0000000e+00],
+       [-2.5147417e-01,  1.0867463e-03,  9.6786356e-01,  0.0000000e+00],
+       [ 4.1018680e-02, -9.9908894e-01,  1.1779292e-02,  1.1794727e+00],
+       [ 0.0000000e+00,  0.0000000e+00,  0.0000000e+00,  1.0000000e+00]])
+        T_velo_2_cam = np.linalg.inv(cam_pose)
+
+        # compute the 3D-2D mapping
+        projected_pix, fov_mask, pix_z = vox2pix(
+            T_velo_2_cam,
+            cam_k,
+            vox_origin,
+            voxel_size * scale_3d,
+            img_W,
+            img_H,
+            scene_size,
+        )
+
+        data["projected_pix_{}".format(scale_3d)] = projected_pix
+        data["pix_z_{}".format(scale_3d)] = pix_z
+        data["fov_mask_{}".format(scale_3d)] = fov_mask
+    return data
 
 def predict(img):
+    img_W, img_H = 640, 480
     img = np.array(img, dtype=np.float32, copy=False) / 255.0
 
     normalize_rgb = transforms.Compose(