def reconstruct_ann(dic):
result = []
for obj in dic:
ratio = COCO.compute_ratio(obj)
obj["ratio"] = ratio
result.append(obj)
return result
def main(input_path, output_path, show=False):
coco = COCO(input_path)
if show:
data_dir = "datasets/coco/train2017/"
dicts = DatasetCatalog.get("coco_2017_train")
metadata = MetadataCatalog.get("coco_2017_train")
for dic in dicts:
for ann in dic["annotations"]:
coco.compute_rbox(ann)
ann["bbox"] = ann["rbox"]
ann["bbox_mode"] = ann["rbox_mode"]
print(ann["bbox"])
image_path = dic["file_name"]
img = utils.convert_PIL_to_numpy(
Image.open(image_path), "RGB")
visualizer = Visualizer(img, metadata=metadata, scale=1)
vis = visualizer.draw_dataset_dict(dic)
webcv2.imshow(image_path+"bbox", vis.get_image()[:, :, ::-1])
webcv2.waitKey()
skip = False
if "val" in input_path:
skip = True
anns = []
for ann_id, ann in tqdm.tqdm(coco.anns.items()):
if skip and ann["iscrowd"]:
continue
ann = coco.compute_rbox(ann)
ann["bbox"] = ann["rbox"]
anns.append(ann)
info = dict(date_created=str(datetime.datetime.now()),
description="Rbox version of {}.".format(
input_path))
coco_dict = dict(
info=info,
categories=coco.dataset["categories"],
annotations=anns,
images=coco.dataset["images"],
license=None)
with open(output_path, "wt") as writer:
json.dump(coco_dict, writer)
def main(input_path, coco_json_path, output_path):
origin_dict = json.load(smart_path(input_path).open("rt"))
coco_dict = json.load(smart_path(coco_json_path).open("rt"))
origin_coco = COCO(input_path)
coco = COCO(coco_json_path)
anns = []
images = list()
image_ids = set()
for ann in origin_dict["annotations"]:
ratio = COCO.compute_ratio(ann)["ratio"]
if ratio > 1 / 3:
continue
ann["ratio"] = ratio
ann["id"] = ann["id"] + 1 << 31
image_id = ann["image_id"] + 1 << 31
image = origin_coco.imgs[ann["image_id"]]
image["id"] = image_id
ann["image_id"] = image_id
anns.append(ann)
if not image_id in image_ids:
images.append(image)
image_ids.add(image_id)
info = dict(date_created=str(datetime.datetime.now()),
description="Merged from {} and {}.".format(
input_path, coco_json_path))
new_coco_dict = dict(info=copy.deepcopy(info),
categories=copy.deepcopy(coco_dict["categories"]),
licenses=None,
annotations=copy.deepcopy(coco_dict["annotations"]) +
copy.deepcopy(anns),
images=copy.deepcopy(coco_dict["images"]) +
copy.deepcopy(images))
with open(output_path, "wt") as writer:
json.dump(new_coco_dict, writer)
def PlotPiecewiseBars(dataset, num_subfigs, sorted=False):
ratios = {
"0-1/5": [0, 1 / 5],
"1/5-1/3": [1 / 5, 1 / 3],
"1/3-1": [1 / 3, 1]
}
dicts = list(DatasetCatalog.get(dataset))
metadata = MetadataCatalog.get(dataset)
labels = metadata.thing_classes
num_thing_per_fig = len(labels) // num_subfigs
bars = dict()
for key in ratios.keys():
bars[key] = [0 for _ in range(len(labels))]
all_ratios = []
for dic in tqdm(dicts):
for obj in dic["annotations"]:
ratio = COCO.compute_ratio(obj, oriented=True)["ratio"]
for key, ratio_range in ratios.items():
if between(ratio, ratio_range):
bars[key][obj["category_id"]] += 1
all_ratios.append(ratio)
slender_ratios = [0 for _ in range(len(labels))]
for i in range(len(labels)):
slender_ratios[i] = bars["0-1/5"][i] / \
(bars["0-1/5"][i]+bars["1/5-1/3"][i]+bars["1/3-1"][i])
slender_ratios = np.array(slender_ratios)
if sorted == True:
sorted_indexes = np.argsort(slender_ratios)
fig, axes = plt.subplots(num_subfigs, 1)
for fig_i in range(num_subfigs):
if sorted == True:
index_range = sorted_indexes[num_thing_per_fig *
fig_i:num_thing_per_fig * (fig_i + 1)]
else:
index_range = np.arrange(num_thing_per_fig * fig_i,
num_thing_per_fig * (fig_i + 1))
label = []
for ind_i in index_range:
label.append(labels[ind_i])
axes[fig_i].set_yscale("symlog")
prev = np.zeros((len(label), ))
for key, bar in bars.items():
bar = np.array(bar)
axes[fig_i].bar(label, bar[index_range], bottom=prev, label=key)
prev = prev + bar[index_range]
axes[fig_i].legend()
fig.set_size_inches(15, 15)
axes[fig_i].set_xticklabels(label, rotation="vertical")
plt.savefig("barchart.jpg")
def repeat_factors_from_ratios(dataset_dicts):
rep_factors = []
for dataset_dict in dataset_dicts:
rep_factor = 0.1
for ann in dataset_dict["annotations"]:
ratio = COCO.compute_ratio(ann)["ratio"]
if ratio < 1 / 5:
rep_factor = 1
break
if ratio < 1 / 3:
rep_factor = 0.5
rep_factors.append(rep_factor)
return torch.tensor(rep_factors, dtype=torch.float32)
def PlotPie(dataset, ratios):
dicts = list(DatasetCatalog.get(dataset))
metadata = MetadataCatalog.get(dataset)
labels = metadata.thing_classes
ratios_num = dict()
for key in ratios.keys():
ratios_num[key] = 0
for dic in tqdm(dicts):
for obj in dic["annotations"]:
ratio = COCO.compute_ratio(obj, oriented=True)["ratio"]
for key, ratio_range in ratios.items():
if between(ratio, ratio_range):
ratios_num[key] += 1
plt.pie(ratios_num.values(), labels=ratios_num.keys(), autopct='%1.2f%%')
plt.savefig("PieChart.jpg")
return ratios_num
all_count = 0
empty_count = 0
problematic_count = 0
for dic in tqdm.tqdm(dicts):
for obj in dic["annotations"]:
all_count += 1
if not "segmentation" in obj or len(obj["segmentation"]) == 0:
empty_count += 1
continue
try:
segmentations = PolygonMasks([obj["segmentation"]])
ratio = segmentations.get_ratios(oriented=True)[0]
print(ratio)
except:
problematic_count += 1
ratio = COCO.compute_ratio(obj)
img = utils.convert_PIL_to_numpy(
Image.open(
io.BytesIO(
smart_path("s3://detection/" +
dic["file_name"]).read_bytes())), "RGB")
visualizer = Visualizer(img, metadata=metadata, scale=scale)
vis = visualizer.draw_dataset_dict(dic)
output(vis, os.path.basename(dic["file_name"]))
print("all", all_count)
print("empty", empty_count)
print("problematic", problematic_count)
def PlotAll(dataset):
dicts = list(DatasetCatalog.get(dataset))
metadata = MetadataCatalog.get(dataset)
labels = metadata.thing_classes
ratios = {
"0-1/5": [0, 1 / 5],
"1/5-1/3": [1 / 5, 1 / 3],
"1/3-1": [1 / 3, 1]
}
bars = dict()
for key in ratios.keys():
bars[key] = [0 for _ in range(len(labels))]
all_ratios = []
ratio_counts = {k: 0 for k in ratios.keys()}
for dic in tqdm(dicts):
for obj in dic["annotations"]:
ratio = COCO.compute_ratio(obj, oriented=True)["ratio"]
for key, ratio_range in ratios.items():
if between(ratio, ratio_range):
bars[key][obj["category_id"]] += 1
ratio_counts[key] += 1
all_ratios.append(ratio)
print("images", len(dicts))
print("counts", ratio_counts)
fig, ax = plt.subplots()
ax.set_yscale("symlog")
prev = np.zeros((len(labels), ))
for key, bar in bars.items():
bar = np.array(bar)
ax.bar(labels, bar, bottom=prev, label=key)
prev = prev + bar
ax.legend()
fig.set_size_inches(18.5, 10.5)
ax.set_xticklabels(labels, rotation="vertical")
group = fig2image(fig)
# cv2.imwrite('./group.png', group)
webcv2.imshow("group", group)
fig, ax = plt.subplots()
all_ratios = sorted(all_ratios)
numbers = [0]
tick = 0.01
seperator = tick
count = 0
for r in all_ratios:
count += 1
while seperator < r:
seperator += tick
numbers.append(count)
ax.plot(np.arange(0, 1, tick), np.array(numbers) / count)
ax.set_xlabel("slenderness")
ax.set_title("cumulative distribution function")
number = fig2image(fig)
# cv2.imwrite('./number.png', number)
webcv2.imshow("number", number)
webcv2.waitKey()
def PlotCOCOAndCOCOPlus(datasets,
num_subfigs,
cmap,
sorted=False,
include_all=False,
show_part=True):
ratios = {
"0-1/5": [0, 1 / 5],
"1/5-1/3": [1 / 5, 1 / 3],
"1/3-1": [1 / 3, 1]
}
total_slender_ratios = []
total_num_all_ratios = []
total_sorted_indexes = []
total_labels = []
min_ratios = []
max_ratios = []
symbols = ["", "+"]
for dataset_i, dataset in enumerate(datasets):
dicts = list(DatasetCatalog.get(dataset))
metadata = MetadataCatalog.get(dataset)
labels = metadata.thing_classes
for i in range(len(labels)):
labels[i] = labels[i] + symbols[dataset_i]
bars = dict()
for key in ratios.keys():
bars[key] = [0 for _ in range(len(labels))]
for dic in tqdm(dicts):
for obj in dic["annotations"]:
ratio = COCO.compute_ratio(obj, oriented=True)["ratio"]
for key, ratio_range in ratios.items():
if between(ratio, ratio_range):
bars[key][obj["category_id"]] += 1
slender_ratios = np.zeros(len(labels))
num_all_ratios = np.zeros(len(labels))
num_slender_ratios = np.array(bars["0-1/5"])
for i in range(len(labels)):
slender_ratios[i] = bars["0-1/5"][i] / \
(bars["0-1/5"][i]+bars["1/5-1/3"][i]+bars["1/3-1"][i])
for key, bar in bars.items():
num_all_ratios = num_all_ratios + np.array(bar)
slender_ratio_all = num_slender_ratios.sum() / num_all_ratios.sum()
if sorted == True:
sorted_indexes = np.argsort(slender_ratios)
min_ratio = slender_ratios[sorted_indexes[0]]
max_ratio = slender_ratios[sorted_indexes[-1]]
else:
sorted_indexes = np.arange(len(labels))
min_ratio = slender_ratios.min()
max_ratio = slender_ratios.max()
if include_all == True:
slender_ratios = np.append(slender_ratios, slender_ratio_all)
num_all_ratios = np.append(num_all_ratios, num_all_ratios.sum())
sorted_indexes = np.insert(sorted_indexes, 0, len(labels))
labels.append('all' + symbols[dataset_i])
total_slender_ratios.append(slender_ratios)
total_num_all_ratios.append(num_all_ratios)
total_sorted_indexes.append(sorted_indexes)
total_labels.append(labels)
min_ratios.append(min_ratio)
max_ratios.append(max_ratio)
if show_part:
show_num = 10
show_indexes = np.append(np.arange(0, show_num + 1),
np.arange(-show_num, 0))
else:
show_indexes = np.arrange(0, len(total_labels[0]))
min_ratio = min(min_ratios)
max_ratio = max(max_ratios)
fig, axes = plt.subplots(num_subfigs, 1)
norm = mpl.colors.Normalize(vmin=min_ratio, vmax=max_ratio)
subplots_adjust(left=0.15,
bottom=0.1,
right=0.9,
top=0.9,
wspace=0,
hspace=0.6)
num_thing_per_fig = len(show_indexes) // num_subfigs
first_row_plus_num = len(show_indexes) % num_subfigs
show_label = []
show_slender_ratio = []
show_num_all_ratio = []
start = 0
end = 0
show_x = []
for fig_i in range(num_subfigs):
start = end
end += num_thing_per_fig
if fig_i == 0:
end += first_row_plus_num
index_range = total_sorted_indexes[0][show_indexes[start:end]]
label = []
slender_ratio = []
num_all_ratio = []
coor_x = []
count = 0
for ind_i in index_range:
for dataset_i in range(len(datasets)):
label.append(total_labels[dataset_i][ind_i])
slender_ratio.append(total_slender_ratios[dataset_i][ind_i])
num_all_ratio.append(total_num_all_ratios[dataset_i][ind_i])
coor_x.append(count)
c = 1.5 if dataset_i == len(datasets) - 1 else 1
count += c
show_label.append(label)
show_slender_ratio.append(slender_ratio)
show_num_all_ratio.append(num_all_ratio)
show_x.append(coor_x)
plt.tick_params(labelsize=16)
bar_ec = [(0.73, 0.73, 0.73, 1.0), (0.5, 0.5, 0.5, 0.8)]
for fig_i in range(num_subfigs):
axes[fig_i].set_yscale("symlog")
prev = np.zeros((len(show_label[fig_i]), ))
bar_colors = [cmap(norm(x)) for x in show_slender_ratio[fig_i]]
bars = axes[fig_i].bar(show_x[fig_i],
show_num_all_ratio[fig_i],
bottom=prev,
color=bar_colors,
edgecolor=bar_ec[fig_i],
width=1)
for i, bar in enumerate(bars):
if i % 2 != 0:
bar.set_hatch('\\')
xticks = show_x[fig_i][::2] + np.ones(len(show_x[fig_i][::2])) * 0.5
axes[fig_i].set_xticks(xticks)
if fig_i == 1:
axes[fig_i].xaxis.tick_top()
axes[fig_i].set_xticklabels(show_label[fig_i][::2],
fontdict={'fontsize': 14},
rotation=45)
else:
axes[fig_i].set_xticklabels(show_label[fig_i][::2],
fontdict={'fontsize': 14},
rotation=315)
legend_labels = ['COCO', 'COCO+']
hatches = ['x', '\\']
patches = []
patches.append(mpatches.Patch(fc='white', label="COCO", ec="grey"))
patches.append(
mpatches.Patch(fc='white', label="COCO+", ec="grey", hatch='\\'))
if fig_i == 0:
axes[fig_i].legend(loc="upper center",
handles=patches,
prop={'size': 24},
ncol=2)
fig.set_size_inches(12, 12)
position = fig.add_axes([0., 0.1, 0.03, 0.8])
fig.colorbar(mpl.cm.ScalarMappable(norm=norm, cmap=cmap),
cax=position,
orientation='vertical')
plt.savefig("coco_coco+.png")
def PlotGradientColorBars(dataset,
num_subfigs,
cmap,
sorted=False,
include_all=False):
ratios = {
"0-1/5": [0, 1 / 5],
"1/5-1/3": [1 / 5, 1 / 3],
"1/3-1": [1 / 3, 1]
}
dicts = list(DatasetCatalog.get(dataset))
metadata = MetadataCatalog.get(dataset)
labels = metadata.thing_classes
bars = dict()
for key in ratios.keys():
bars[key] = [0 for _ in range(len(labels))]
for dic in tqdm(dicts):
for obj in dic["annotations"]:
ratio = COCO.compute_ratio(obj, oriented=True)["ratio"]
for key, ratio_range in ratios.items():
if between(ratio, ratio_range):
bars[key][obj["category_id"]] += 1
slender_ratios = np.zeros(len(labels))
num_all_ratios = np.zeros(len(labels))
num_slender_ratios = np.array(bars["0-1/5"])
for i in range(len(labels)):
slender_ratios[i] = bars["0-1/5"][i] / \
(bars["0-1/5"][i]+bars["1/5-1/3"][i]+bars["1/3-1"][i])
for key, bar in bars.items():
num_all_ratios = num_all_ratios + np.array(bar)
slender_ratio_all = num_slender_ratios.sum() / num_all_ratios.sum()
if sorted == True:
sorted_indexes = np.argsort(slender_ratios)
min_ratio = slender_ratios[sorted_indexes[0]]
max_ratio = slender_ratios[sorted_indexes[-1]]
else:
sorted_indexes = np.arange(len(labels))
min_ratio = slender_ratios.min()
max_ratio = slender_ratios.max()
if include_all == True:
slender_ratios = np.append(slender_ratios, slender_ratio_all)
num_all_ratios = np.append(num_all_ratios, num_all_ratios.sum())
sorted_indexes = np.insert(sorted_indexes, 0, len(labels))
labels.append('all')
fig, axes = plt.subplots(num_subfigs, 1)
norm = mpl.colors.Normalize(vmin=min_ratio, vmax=max_ratio)
subplots_adjust(left=0.15,
bottom=0.1,
right=0.9,
top=0.9,
wspace=0,
hspace=0.6)
num_thing_per_fig = len(labels) // num_subfigs
for fig_i in range(num_subfigs):
start = num_thing_per_fig * fig_i
end = num_thing_per_fig * (fig_i + 1)
if fig_i == num_subfigs - 1:
end = len(labels)
index_range = sorted_indexes[start:end]
label = []
for ind_i in index_range:
label.append(labels[ind_i])
axes[fig_i].set_yscale("symlog")
prev = np.zeros((len(label), ))
bar_colors = [cmap(norm(x)) for x in slender_ratios[index_range]]
axes[fig_i].bar(label,
num_all_ratios[index_range],
bottom=prev,
color=bar_colors,
edgecolor=(0, 0, 0))
axes[fig_i].set_xticklabels(label, rotation="vertical")
fig.set_size_inches(12, 12)
position = fig.add_axes([0., 0.1, 0.03, 0.8])
fig.colorbar(mpl.cm.ScalarMappable(norm=norm, cmap=cmap),
cax=position,
orientation='vertical')
plt.savefig("barchart.png")