def make_frame(t): wallclock = ((t / duration_sec)**time_warp) * times[-1] png, kimg, lod = snaps[max(bisect.bisect(times, wallclock) - 1, 0)] if png_cache[0] == png: img = png_cache[1] else: img = scipy.misc.imread(png) while img.shape[1] > 1920 or img.shape[0] > 1080: img = (img.astype(np.float32).reshape(img.shape[0] // 2, 2, img.shape[1] // 2, 2, -1).mean(axis=(1, 3))) png_cache[:] = [png, img] img = misc.draw_text_label(img, "lod %.2f" % lod, 16, img.shape[0] - 4, alignx=0.0, aligny=1.0) img = misc.draw_text_label( img, misc.format_time(int(np.rint(wallclock))), img.shape[1] // 2, img.shape[0] - 4, alignx=0.5, aligny=1.0, ) img = misc.draw_text_label( img, "%.0f kimg" % kimg, img.shape[1] - 16, img.shape[0] - 4, alignx=1.0, aligny=1.0, ) return img
def make_frame(t): wallclock = ((t / duration_sec) ** time_warp) * times[-1] png, kimg, lod = snaps[max(bisect.bisect(times, wallclock) - 1, 0)] if png_cache[0] == png: img = png_cache[1] else: img = scipy.misc.imread(png) while img.shape[1] > 1920 or img.shape[0] > 1080: img = img.astype(np.float32).reshape(img.shape[0]//2, 2, img.shape[1]//2, 2, -1).mean(axis=(1,3)) png_cache[:] = [png, img] img = misc.draw_text_label(img, 'lod %.2f' % lod, 16, img.shape[0]-4, alignx=0.0, aligny=1.0) img = misc.draw_text_label(img, misc.format_time(int(np.rint(wallclock))), img.shape[1]//2, img.shape[0]-4, alignx=0.5, aligny=1.0) img = misc.draw_text_label(img, '%.0f kimg' % kimg, img.shape[1]-16, img.shape[0]-4, alignx=1.0, aligny=1.0) return img
def make_frame(t): wallclock = ((t / duration_sec) ** time_warp) * times[-1] png, kimg, lod = snaps[max(bisect.bisect(times, wallclock) - 1, 0)] if png_cache[0] == png: img = png_cache[1] else: img = scipy.misc.imread(png) # Check if the img is greyscale. If so, it needs to be converted to RGB, since MoviePy expects it in that format if img.ndim == 2: tmpimg = img img = np.zeros([tmpimg.shape[0], tmpimg.shape[1], 3], tmpimg.dtype) img[:,:,0] = tmpimg img[:,:,1] = tmpimg img[:,:,2] = tmpimg while img.shape[1] > 1920 or img.shape[0] > 1080: img = img.astype(np.float32).reshape(img.shape[0]//2, 2, img.shape[1]//2, 2, -1).mean(axis=(1,3)) png_cache[:] = [png, img] img = misc.draw_text_label(img, 'lod %.2f' % lod, 16, img.shape[0]-4, alignx=0.0, aligny=1.0) img = misc.draw_text_label(img, misc.format_time(int(np.rint(wallclock))), img.shape[1]//2, img.shape[0]-4, alignx=0.5, aligny=1.0) img = misc.draw_text_label(img, '%.0f kimg' % kimg, img.shape[1]-16, img.shape[0]-4, alignx=1.0, aligny=1.0) return img
def evaluate_metrics(run_id, log, metrics, num_images, real_passes, minibatch_size=None): metric_class_names = { 'swd': 'metrics.sliced_wasserstein.API', 'fid': 'metrics.frechet_inception_distance.API', 'is': 'metrics.inception_score.API', 'msssim': 'metrics.ms_ssim.API', } # Locate training run and initialize logging. result_subdir = misc.locate_result_subdir(run_id) snapshot_pkls = misc.list_network_pkls(result_subdir, include_final=False) assert len(snapshot_pkls) >= 1 log_file = os.path.join(result_subdir, log) print('Logging output to', log_file) misc.set_output_log_file(log_file) # Initialize dataset and select minibatch size. dataset_obj, mirror_augment = misc.load_dataset_for_previous_run(result_subdir, verbose=True, shuffle_mb=0) if minibatch_size is None: minibatch_size = np.clip(8192 // dataset_obj.shape[1], 4, 256) # Initialize metrics. metric_objs = [] for name in metrics: class_name = metric_class_names.get(name, name) print('Initializing %s...' % class_name) class_def = tfutil.import_obj(class_name) image_shape = [3] + dataset_obj.shape[1:] obj = class_def(num_images=num_images, image_shape=image_shape, image_dtype=np.uint8, minibatch_size=minibatch_size) tfutil.init_uninited_vars() mode = 'warmup' obj.begin(mode) for idx in range(10): obj.feed(mode, np.random.randint(0, 256, size=[minibatch_size]+image_shape, dtype=np.uint8)) obj.end(mode) metric_objs.append(obj) # Print table header. print() print('%-10s%-12s' % ('Snapshot', 'Time_eval'), end='') for obj in metric_objs: for name, fmt in zip(obj.get_metric_names(), obj.get_metric_formatting()): print('%-*s' % (len(fmt % 0), name), end='') print() print('%-10s%-12s' % ('---', '---'), end='') for obj in metric_objs: for fmt in obj.get_metric_formatting(): print('%-*s' % (len(fmt % 0), '---'), end='') print() # Feed in reals. for title, mode in [('Reals', 'reals'), ('Reals2', 'fakes')][:real_passes]: print('%-10s' % title, end='') time_begin = time.time() labels = np.zeros([num_images, dataset_obj.label_size], dtype=np.float32) [obj.begin(mode) for obj in metric_objs] for begin in range(0, num_images, minibatch_size): end = min(begin + minibatch_size, num_images) images, labels[begin:end] = dataset_obj.get_minibatch_np(end - begin) if mirror_augment: images = misc.apply_mirror_augment(images) if images.shape[1] == 1: images = np.tile(images, [1, 3, 1, 1]) # grayscale => RGB [obj.feed(mode, images) for obj in metric_objs] results = [obj.end(mode) for obj in metric_objs] print('%-12s' % misc.format_time(time.time() - time_begin), end='') for obj, vals in zip(metric_objs, results): for val, fmt in zip(vals, obj.get_metric_formatting()): print(fmt % val, end='') print() # Evaluate each network snapshot. for snapshot_idx, snapshot_pkl in enumerate(reversed(snapshot_pkls)): prefix = 'network-snapshot-'; postfix = '.pkl' snapshot_name = os.path.basename(snapshot_pkl) assert snapshot_name.startswith(prefix) and snapshot_name.endswith(postfix) snapshot_kimg = int(snapshot_name[len(prefix) : -len(postfix)]) print('%-10d' % snapshot_kimg, end='') mode ='fakes' [obj.begin(mode) for obj in metric_objs] time_begin = time.time() with tf.Graph().as_default(), tfutil.create_session(config.tf_config).as_default(): G, D, Gs = misc.load_pkl(snapshot_pkl) for begin in range(0, num_images, minibatch_size): end = min(begin + minibatch_size, num_images) latents = misc.random_latents(end - begin, Gs) images = Gs.run(latents, labels[begin:end], num_gpus=config.num_gpus, out_mul=127.5, out_add=127.5, out_dtype=np.uint8) if images.shape[1] == 1: images = np.tile(images, [1, 3, 1, 1]) # grayscale => RGB [obj.feed(mode, images) for obj in metric_objs] results = [obj.end(mode) for obj in metric_objs] print('%-12s' % misc.format_time(time.time() - time_begin), end='') for obj, vals in zip(metric_objs, results): for val, fmt in zip(vals, obj.get_metric_formatting()): print(fmt % val, end='') print() print()
def compute_fid(Gs, minibatch_size, dataset_obj, iter_number, lod=0, num_images=10000, printing=True): # Initialize metrics. from metrics.frechet_inception_distance import API as class_def image_shape = [3] + dataset_obj.shape[1:] obj = class_def(num_images=num_images, image_shape=image_shape, image_dtype=np.uint8, minibatch_size=minibatch_size) mode = 'warmup' obj.begin(mode) for idx in range(10): obj.feed( mode, np.random.randint(0, 256, size=[minibatch_size] + image_shape, dtype=np.uint8)) obj.end(mode) # Print table header if printing: print(flush=True) print('%-10s%-12s' % ('KIMG', 'Time_eval'), end='', flush=True) print('%-12s' % ('FID'), end='', flush=True) print(flush=True) print('%-10s%-12s%-12s' % ('---', '---', '---'), end='', flush=True) print(flush=True) # Feed in reals. print('%-10s' % "Reals", end='', flush=True) time_begin = time.time() labels = np.zeros([num_images, dataset_obj.label_size], dtype=np.float32) obj.begin(mode) for begin in range(0, num_images, minibatch_size): end = min(begin + minibatch_size, num_images) images, labels[begin:end] = dataset_obj.get_minibatch_np(end - begin, lod=lod) if images.shape[1] == 1: images = np.tile(images, [1, 3, 1, 1]) # grayscale => RGB obj.feed(mode, images) results = obj.end(mode) if printing: print('%-12s' % misc.format_time(time.time() - time_begin), end='', flush=True) print(results[0], end='', flush=True) print(flush=True) # Evaluate each network snapshot. if printing: print('%-10d' % iter_number, end='', flush=True) mode = 'fakes' obj.begin(mode) time_begin = time.time() for begin in range(0, num_images, minibatch_size): end = min(begin + minibatch_size, num_images) latents = misc.random_latents(end - begin, Gs) images = Gs.run(latents, labels[begin:end], num_gpus=config.num_gpus, out_mul=127.5, out_add=127.5, out_dtype=np.uint8) if images.shape[1] == 1: images = np.tile(images, [1, 3, 1, 1]) # grayscale => RGB obj.feed(mode, images) results = obj.end(mode) if printing: print('%-12s' % misc.format_time(time.time() - time_begin), end='', flush=True) print(results[0], end='', flush=True) print(flush=True) return results[0]