upload joint calibration data

others/joint_calib/greenhead/README.md

@@ -0,0 +1,22 @@
+## Joint calibration
+
+### What's this?
+
+Recall that we have two sets of cameras in the dataset, DSLR (48) and high speed (24) cameras. These two sets of cameras have different color spaces, thus the images cannot be used together. For example, in the inverse rendering experiments in Section 4.1 and 4.3, we only use the images captured by the DSLR cameras, whereas in Section 4.2, we only use the images captured by the high speed cameras. As a result, we didn't perform joint camera calibration on these two sets of images in the main paper because the color differences makes the COLMAP hard to work. Even if we can perform some color postprocessing, e.g. Photoshop, the resulting images are not unreliable enough to run inverse rendering algorithms.
+
+However, one may still need this joint calibration results. For example, let's say we want to perform linear combination on the OLAT images, which requires the light positions relative to each high speed camera. But in the main repo, the light positions are relative to the DSLR cameras. In this case, we need to calibrate the two sets of cameras jointly to transform the light positions from DSLR coordinate system to high speed camera coordinate system.
+
+In practice, we manually correct the color of the high speed camera images with the DSLR images in Photoshop and then run COLMAP to register all the cameras.
+
+### File format
+#### images
+There are 71 images here. The image colors have been manually corrected using Photoshop. One image is missed since the corresponding camera was down at the capturing time.
+
+#### mappings.txt
+The images have different names (0000.jpg,0001.jpg,...) from the images in the main repo (CA2.jpg,...). This file contains the mapping between the two sets of names.
+
+#### cameras.json
+This json file contains the results of COLMAP. K and dist are the intrinsics, newK is the undistorted intrinsics. We only provide this results for reference, you may rerun COLMAP by yourself.
+
+#### cameras_normalized.json
+This json file contains the same content as cameras.json, except that the camera poses are normalized to match the poses in the main repo.