| import os |
| import mat73 |
| import torch |
| import numpy as np |
| import json |
| import h5py |
|
|
| def read_h5(path): |
| with h5py.File(path, 'r') as f: |
| frames = np.array(f['data']) |
| return frames |
|
|
| def process_captured_data(files_path, savepath): |
| for file in os.listdir(files_path): |
| if file[-3:] == "mat" and (file[:-3]+"pt") not in os.listdir(savepath): |
| filepath = os.path.join(files_path, file) |
| rgba = mat73.loadmat(filepath)["transient"].transpose(1, 2, 0)[..., :][..., None] |
| rgba = np.flip(np.flip(rgba, 0), 1) |
| for i in range(3): |
| rgba = (rgba[1::2, ::2] + rgba[::2, ::2] + rgba[1::2, 1::2] + rgba[::2, 1::2])/4 |
| torch.save(rgba, os.path.join(savepath, file[:-3]+"pt")) |
|
|
| def bundle_rays(pathToH5s, outputPath, trainJsonPath): |
| with open(trainJsonPath, "r") as fp: |
| meta = json.load(fp) |
| train_fnames = [] |
| for i in range(len(meta["frames"])): |
| frame = meta["frames"][i] |
| fname = frame["filepath"][:-2]+"h5" |
| train_fnames.append(fname) |
|
|
| frames = read_h5(os.path.join(pathToH5s, train_fnames[0])) |
| w = frames.shape[0] |
| h = frames.shape[1] |
| bins = frames.shape[2] |
| |
| x = np.linspace(0, h-1, h) |
| y = np.linspace(0, w-1, w) |
| X, Y = np.meshgrid(x, y) |
|
|
| if len(frames.shape) == 4: |
| channels = 3 |
| else: |
| channels = 1 |
| num_train_files = len(train_fnames) |
| |
| data_array = np.zeros((w*h*num_train_files, bins, channels), dtype=np.float32) |
| x_array = np.zeros(w*h*num_train_files) |
| y_array = np.zeros(w*h*num_train_files) |
| file_prefix_array = np.zeros(w*h*num_train_files) |
|
|
| for ind, file in enumerate(train_fnames): |
| print("Opening: " + file) |
| frames = read_h5(os.path.join(pathToH5s, file)) |
| frames = frames.reshape(-1, frames.shape[2], frames.shape[3]) |
| |
| data_array[ind*w*h:(ind+1)*w*h] = frames[..., :3] |
| x_array[ind*w*h:(ind+1)*w*h] = X.flatten() |
| y_array[ind*w*h:(ind+1)*w*h] = Y.flatten() |
| file_prefix_array[ind*w*h:(ind+1)*w*h] = ind |
| |
| p = np.random.permutation(data_array.shape[0]) |
| data_array = data_array[p] |
| x_array = x_array[p] |
| y_array = y_array[p] |
| file_prefix_array = file_prefix_array[p] |
|
|
|
|
| print("Outputting to files") |
| file = h5py.File(os.path.join(outputPath, "samples.h5"), 'w') |
| dataset = file.create_dataset( |
| "dataset", data_array.shape, dtype='f', data=data_array |
| ) |
| file.close() |
|
|
| file = h5py.File(os.path.join(outputPath, "x.h5"), 'w') |
| dataset = file.create_dataset( |
| "dataset", x_array.shape, dtype='f', data=x_array |
| ) |
| file.close() |
|
|
| file = h5py.File(os.path.join(outputPath, "y.h5"), 'w') |
| dataset = file.create_dataset( |
| "dataset", y_array.shape, dtype='f', data=y_array |
| ) |
| file.close() |
| file = h5py.File(os.path.join(outputPath, "file_indices.h5"), 'w') |
| dataset = file.create_dataset( |
| "dataset", file_prefix_array.shape, dtype='f', data=file_prefix_array |
| ) |
| file.close() |
|
|
|
|
| def bundle_rays_cap(pathToH5s, outputPath, trainJsonPath): |
| with open(trainJsonPath, "r") as fp: |
| meta = json.load(fp) |
| train_fnames = [] |
| for i in range(len(meta["frames"])): |
| frame = meta["frames"][i] |
| fname = frame["filepath"][:-2].split("/")[-1]+"mat" |
| train_fnames.append(fname) |
|
|
| |
| frames = mat73.loadmat(os.path.join(pathToH5s, train_fnames[0]))["transient"].transpose(1, 2, 0) |
| w = frames.shape[0] |
| h = frames.shape[1] |
| |
| bins = 3000 |
| |
| x = np.linspace(0, h-1, h) |
| y = np.linspace(0, w-1, w) |
| X, Y = np.meshgrid(x, y) |
|
|
| |
| |
| |
| channels = 1 |
| num_train_files = len(train_fnames) |
| |
| data_array = np.zeros((w*h*num_train_files, bins, channels), dtype=np.float32) |
| x_array = np.zeros(w*h*num_train_files) |
| y_array = np.zeros(w*h*num_train_files) |
| file_prefix_array = np.zeros(w*h*num_train_files) |
|
|
| for ind, file in enumerate(train_fnames): |
| print("Opening: " + file) |
| |
| frames = mat73.loadmat(os.path.join(pathToH5s, file))["transient"].transpose(1, 2, 0) |
| frames = frames.reshape(-1, frames.shape[2]) |
| |
| data_array[ind*w*h:(ind+1)*w*h] = frames[..., :bins, None] |
| |
| x_array[ind*w*h:(ind+1)*w*h] = X.flatten() |
| y_array[ind*w*h:(ind+1)*w*h] = Y.flatten() |
| file_prefix_array[ind*w*h:(ind+1)*w*h] = ind |
| |
| p = np.random.permutation(data_array.shape[0]) |
| data_array = data_array[p] |
| x_array = x_array[p] |
| y_array = y_array[p] |
| file_prefix_array = file_prefix_array[p] |
|
|
|
|
| print("Outputting to files") |
| file = h5py.File(os.path.join(outputPath, "samples.h5"), 'w') |
| dataset = file.create_dataset( |
| "dataset", data_array.shape, dtype='f', data=data_array |
| ) |
| file.close() |
|
|
| file = h5py.File(os.path.join(outputPath, "x.h5"), 'w') |
| dataset = file.create_dataset( |
| "dataset", x_array.shape, dtype='f', data=x_array |
| ) |
| file.close() |
|
|
| file = h5py.File(os.path.join(outputPath, "y.h5"), 'w') |
| dataset = file.create_dataset( |
| "dataset", y_array.shape, dtype='f', data=y_array |
| ) |
| file.close() |
| file = h5py.File(os.path.join(outputPath, "file_indices.h5"), 'w') |
| dataset = file.create_dataset( |
| "dataset", file_prefix_array.shape, dtype='f', data=file_prefix_array |
| ) |
| file.close() |
| |
| if __name__=="__main__": |
| pass |
