PASCAL Visual Object Classes Challenge 2011 (VOC2011) 图像标注指南标注规则

admin管理员组

文章数量:1576265

PASCAL Visual Object Classes Challenge 2011 {VOC2011} Annotation Guidelines

• 1. Guidelines on what and how to label
• 2. Guidelines on categorisation
• 3. Guidelines on segmentation
• 4. 扩展知识
• Wordbook
• References

http://host.robots.ox.ac.uk/pascal/VOC/voc2012/guidelines.html

1. Guidelines on what and how to label

1.1. What to label

All objects of the defined categories, unless:

• you are unsure what the object is.
• the object is very small (at your discretion).
• less than 10-20% of the object is visible, such that you cannot be sure what class it is. e.g. if only a tyre is visible it may belong to car or truck so cannot be labelled car, but feet / faces can only belong to a person.

If this is not possible because too many objects, mark image as bad.
如果图片内物体过多导致无法标注，标记为 bad。

discretion [dɪˈskreʃn]：n. 自由裁量权，谨慎，判断力，判定，考虑周到
tyre [ˈtaɪə(r)]：n. 轮胎，轮箍 v. 装轮胎于...
feet [fiːt]：n. 脚 (foot 的复数形式)，尺，韵脚
category [ˈkætəɡəri]：n. 种类，分类，范畴

1.2. Viewpoint

Record the viewpoint of the ‘bulk’ of the object e.g. the body rather than the head. Allow viewpoints within 10-20 degrees.
If ambiguous, leave as ‘Unspecified’. Unusually rotated objects e.g. upside-down people should be left as ‘Unspecified’.
图片中物体相对 camera 的角度，理论上有用，但实际上物体也会自己形变转动，感觉意义有限。

upside-down ['ʌp,saɪd'daʊn]：颠倒的，乱七八糟的
ambiguous [æmˈbɪɡjuəs]：adj. 模糊不清的，模棱两可的，不明确的，不明朗的，引起歧义的
bulk [bʌlk]：n. 体积，容量，大多数，大部分，大块 vt. 使扩大，使形成大量，使显得重要
head [hed]：n. 头，头痛，上端，最前的部分，理解力 vt. 前进，用头顶，作为...的首领，站在...的前头，给...加标题 vi. 出发，成头状物，船驶往 adj. 头的，主要的，在顶端的
unspecified [ʌnˈspesɪfaɪd]：adj. 未指明的，未详细说明的
degree [dɪˈɡriː]：n. 程度，等级，度，学位，阶层

1.3. Bounding box

Mark the bounding box of the visible area of the object (not the estimated total extent of the object).
Bounding box should contain all visible pixels, except where the bounding box would have to be made excessively large to include a few additional pixels (< 5%) e.g. a car aerial.
检测框需要包含物体全部像素，除非物体有过于细长伸出的臂 (伸出部分占像素 < 5%)。

extent [ɪkˈstent]：n. 程度，范围，长度
excessively [ɪkˈsesɪvli]：adv. 过分地，极度
aerial [ˈeəriəl]：adj. 空中的，航空的，空气的，空想的

1.4. Truncation

If more than 15-20% of the object lies outside the bounding box mark as Truncated. The flag indicates that the bounding box does not cover the total extent of the object.
如果物体超过 15-20% 的区域都不在标注框内，则备注为 Truncated.

truncation [trʌŋˈkeɪʃn]：n. 截断，切掉顶端
truncate [trʌŋˈkeɪt]：v. 截断，删节，把...截成平面 adj. (羽毛等) 截形的，截短的
occlude [ə'kluːd]：v. 使闭塞，堵塞

1.5. Occlusion

If more than 5% of the object is occluded within the bounding box, mark as Occluded. The flag indicates that the object is not totally visible within the bounding box.
如果框内物体超过 5% 的区域被遮挡，则标注为 occluded.

1.6. Image quality / illumination

Images which are poor quality (e.g. excessive motion blur) should be marked bad. However, poor illumination (e.g. objects in silhouette) should not count as poor quality unless objects cannot be recognised.
对于图片质量差，比如过多运动模糊，过强光照，除了滤除坏样本，还可以考虑识别出场景，比如夜晚，或者强阳光情况。

Images made up of multiple images (e.g. collages) should be marked bad.

silhouette [ˌsɪluˈet]：n. 轮廓，剪影 vt. 使...照出影子来，使...仅仅显出轮廓
collage [ˈkɒlɑːʒ]：n. 抽象拚贴画 (用报纸、布、压平的花等碎片拼合而成的) vt. 把...创作成拼贴画，拼贴
illumination [ɪˌluːmɪˈneɪʃn]：n. 照明，照度，启发，灯饰，阐明
recognise ['rekəɡnaɪz]：vt. 认出，承认，认可，识别

1.7. Clothing / mud / snow etc.

If an object is ‘occluded’ by a close-fitting occluder e.g. clothing, mud, snow etc., then the occluder should be treated as part of the object.
如果是物体被衣服，泥或雪这种可穿着的物体部分遮盖，则不算 occluded.

mud [mʌd]：泥，诽谤的话，无价值的东西
close-fitting [ˌkloʊs ˈfɪtɪŋ]：adj. 贴身的，紧身的
occluder：n. 遮光板，咬合器

1.8. Transparency

Do label objects visible through glass, but treat reflections on the glass as occlusion.
标记通过玻璃可见的物体，但将玻璃上的反射视为遮挡。

transparency [trænsˈpærənsi]：n. 透明，透明度，幻灯片，有图案的玻璃
occlusion [ə'klu:ʒn]：n. 闭塞，咬合，锢囚锋，全闭合音

1.9. Mirrors

Do label objects in mirrors.

mirror [ˈmɪrə(r)]：n. 镜子，真实的写照，榜样 vt. 反射，反映

1.10. Pictures

Label objects in pictures / posters / signs only if they are photorealistic but not if cartoons, symbols etc.

poster [ˈpəʊstə(r)]：n. 海报，广告，招贴
sign [saɪn] ：n. 迹象，符号，记号，手势，指示牌 vi. 签署，签名 vt. 签署，示意
photorealistic：adj. 照相写实主义的，超级现实主义的
cartoon [kɑːˈtuːn]：n. 卡通片，动画片，连环漫画 vt. 为...画漫画 vi. 画漫画
symbol [ˈsɪmbl]：n. 象征，符号，标志

2. Guidelines on categorisation

分类

categorisation [,kætɪɡərɪ'zeʃən]：分类  (categorise 的变体)，范畴化 (categorise 的变体)

Aeroplane

Includes gliders but not hang gliders or helicopters

glider [ˈɡlaɪdə(r)]：n. 滑翔机，滑翔员，滑翔导弹
hang [hæŋ]：vt. 悬挂，垂下，装饰，绞死，使悬而未决 vi. 悬着，垂下，被绞死，悬而不决 n. 悬挂，暂停，中止
hang glider：悬挂式滑翔机，滑翔风筝
helicopter ['helɪkɒptə]：直升飞机

Bicycle

Includes tricycles, unicycles

tricycle [‘traɪsɪk(ə)l]：三轮车 
unicycle [‘ju:nisaikl]：独轮脚踏车

Bird

All birds

Boat

Ships, rowing boats, pedaloes but not jet skis

jet ski：摩托艇 
pedalo [‘pedələʊ]：脚踏船
rowing boat：划艇，小划子，游艇
row [rəʊ]：n. 行，排，划船，街道，吵闹 vt. 划船，使...成排 vi. 划船，争吵
jet [dʒet]：n. 喷射，喷嘴，喷气式飞机，黑玉 adj. 墨黑的 vt. 射出 vi. 射出，乘喷气式飞机
ski [skiː]：n. 滑雪橇 vi. 滑雪 adj. 滑雪 (用) 的

Bottle

Plastic, glass or feeding bottles

feeding bottle：奶瓶，哺乳瓶
bottle [ˈbɒtl]：n. 瓶子，一瓶的容量 vt. 控制，把...装入瓶中 vi. (街头艺人演出后) 收拢钱币
plastic [ˈplæstɪk]：n. 塑料，塑料学，(非正式) 信用卡 adj. 塑料制的，人造的，不自然的，(物质、材料) 可塑的，产生立体感的，塑性的，有创造力的，适应环境变化的

Bus

Includes minibus but not trams

Car

Includes cars, vans, large family cars for 6-8 people etc.
Excludes go-carts, tractors, emergency vehicles, lorries / trucks etc.
Do not label where only the vehicle interior is shown.
Include toys that look just like real cars, but not ‘cartoony’ toys.

tram [træm]：n. 电车轨道，煤车 vt. 用煤车运载 vi. 乘电车
vehicle interior：汽车内饰
go-cart ['ɡəukɑ:t]：早期的轻便马车，学走器，手推车，竞赛用的微型单座汽车
lorry ['lɒrɪ]：卡车，货车，运料车
emergency [ɪˈmɜːdʒənsi]：n. 紧急情况，突发事件，非常时刻 adj. 紧急的，备用的
exclude [ɪkˈskluːd]：vt. 排除，排斥，拒绝接纳，逐出
van [væn]：n. 先锋，厢式货车，增值网vt. 用车搬运
tractor [ˈtræktə(r)]：n. 拖拉机，牵引机
toy [tɔɪ]：n. 玩具，小装饰品，不值钱的东西 vi. 玩弄，调情，随随便便地对待 adj. 作为玩具的，玩物似的

Cat

Domestic cats (not lions etc.)

Chair

Includes armchairs, deckchairs but not stools or benches.
Excludes seats in buses, cars etc.
Excludes wheelchairs.

stool [stuːl]：凳子，粪便，厕所
wheelchair ['wiːltʃeə]：轮椅
domestic [də'mestɪk]：n. 佣人，家佣，家庭纠纷，家庭矛盾 adj.本国的，国内的，家用的，家庭的
lion [ˈlaɪən]：n. 狮子，名人，勇猛的人，社交场合的名流
armchair [ˈɑːmtʃeə(r)]：n. 扶手椅，单人沙发 adj. 不切实际的
deckchair [ˈdektʃer]：n. 帆布躺椅
stool [stuːl]：n. 凳子，粪便，厕所 vi. 长新枝，分檗 vt. 引诱，诱捕
bench [bentʃ]：n. 长凳，工作台，替补队员 vt. 给...以席位，为...设置条凳

Cow

All cows

cow [kaʊ]：n. 奶牛，母牛，母兽 vt. 威胁，恐吓

Dining table

Only tables for eating at.
Not coffee tables, desks, side tables or picnic benches

dine [daɪn]：vt. 宴请 vi. 进餐，用餐
picnic [ˈpɪknɪk]：n. 野餐 vi. 去野餐

Dog

Domestic dogs (not wolves etc.)

wolf [wʊlf]：n. 狼，色狼，残忍贪婪之人 vt. 大吃，狼吞虎咽地吃

Horse

Includes ponies, donkeys, mules etc.

donkey [‘dɒŋkɪ]：驴子，傻瓜，顽固的人
pony [ˈpəʊni]：n. 矮种马，小型马 adj. 小型的，每日摘要的 vt. 付清 vi. 付清
mule [mjuːl]：n. 骡，倔强之人，顽固的人，杂交种动物

Motorbike

Includes mopeds, scooters, sidecars

scooter ['skuːtə]：小轮摩托车，速可达，单脚滑行车，小孩滑板车
moped [ˈməʊped]：n. 助动车，机动脚踏两用车 v. 抑郁不乐，生闷气 (mope 的过去式和过去分词)
sidecar [ˈsaɪdkɑː(r)]：n. 轻快的双轮马车，跨斗

People

Includes babies, faces (i.e. truncated people)

Potted plant

Indoor plants excluding flowers in vases, or outdoor plants clearly in a pot.

potted plant：盆栽植物
plant [plɑːnt]：n. 工厂，车间，植物，设备，庄稼 vt. 种植，培养，栽培，安置 vi. 种植
pot [pɒt]：n. 壶，盆，罐 vt. 把...装罐，射击，节略 vi. 随手射击
vase [vɑːz]：n. 瓶，花瓶

Sheep

Sheep, not goats

goat [ɡəʊt]：n. 山羊，替罪羊，色鬼

Sofa

Excludes sofas made up as sofa-beds

Train

Includes train carriages, excludes trams

tram [træm]：电车轨道，煤车
carriage [ˈkærɪdʒ]：n. 运输，运费，四轮马车，举止，客车厢

TV/monitor

Standalone screens (not laptops), not advertising displays
TV/显示器：笔记本电脑不算，电子广告牌不算。

standalone ['stændə,ləʊn]：adj. 独立运行的，独立的 n. 脱机
laptop [ˈlæptɒp]：n. 膝上型轻便电脑，笔记本电脑
advertisement [ədˈvɜ:tɪsmənt]：n. 广告，宣传

3. Guidelines on segmentation

segmentation [ˌseɡmenˈteɪʃn]：n. 分割，割断，细胞分裂

3.1. What to segment

Objects whose bounding boxes have been labelled according to the above guidelines.
You may need to exclude backpacks, handbags etc. which were included in the bounding box.
You may also need to include hands, chair legs etc. which were outside the bounding box.
分割画边界需要去除背包、手提包之类的身外物体，但不要漏掉椅子腿、手臂之类的这种长出去的部位。

backpack [ˈbækpæk]：n. 双肩背包，背包，远足用的背包 vt. 挑运，把...放入背包 vi. 背负简便行李旅行，挑运
segment [ˈseɡmənt]：n. 段，部分，(水果或花的) 瓣，(动物的) 节，弓形，音段，(广播或节目的) 时间段 v. 分割，(细胞) 分裂

3.2. Accuracy

Segment within 5 pixels. Labelled pixels MUST be the object;
pixels outside the 5-pixel border area MUST be background. Border pixels can be either. Use the tri-map displayed by the segmentation tool to ensure these constraints hold.
This may involve labelling pixels outside the bounding box.
分割描边精度有 5 个像素的自由度，保证外部的肯定是背景。

involve [ɪnˈvɒlv]：vt. 包含，牵涉，使陷于，潜心于

3.3. Mixed pixels/ transparency

Pixels which are mixed e.g. due to transparency, motion blur or the presence of a border should be considered to belong to the object whose colour contributes most to the mix.
对混合的像素 (由于运动，透明性等)，如果能确认属于物体，依旧要标注。

transparency [trænsˈpærənsi]：n. 透明，透明度，幻灯片，有图案的玻璃
border [ˈbɔːdə(r)]：n. 边境，边界，国界 vt. 接近，与...接壤，在...上镶边 vi. 接界，近似
presence [ˈprezns]：n. 存在，出席，参加，风度，仪态

3.4. Thin structures

Aim to capture thin structures where possible, within the accuracy constraints. Structures of around one pixel thickness can be ignored e.g. wires, rigging, whiskers.

rigging：索具，绳索；装备，传动装置
whisker ['wɪskə]：晶须；胡须；腮须

3.5. Objects on tables etc.

If a number of small objects are occluding an object e.g. cutlery / silverware on a dining table, they can be considered part of that object. The exception is if they are sticking out of the object (e.g. candles) where they should be truncated at the object boundary.

cutlery ['kʌtlərɪ]：餐具，刀剑制造业
silverware ['sɪlvəweə]：银器，镀银餐具
candle ['kænd(ə)l]：蜡烛，烛光，烛形物
sticker [ˈstɪkə(r)]：n. 粘贴标签，张贴物，坚持不懈的人，尖刀，难题

3.6. Difficult images

Images which are overly difficult to segment to the required accuracy can be left unlabelled e.g. a nest of bicycles.
物体过于难标注，比如一堆自行车。

4. 扩展知识

• 如果按一种固定标准去给出一个布尔值 (True / False)，则会让我们的标注与当前 project 绑定，甚至和算法绑定，这显然并不划算。
• 让标注人员估计物体可见的百分比，会带来极高的标注代价。加之 “However, training humans to visually inspect a bounding box with IOU of 0.3 and distinguish it from one with IOU 0.5 is surprisingly difficult.”，这也会成为误差的来源。一个相对合理的折中，是可以分 3-5 级，来指示物体的 visibility.
• 在检测中我们用矩形框标记物体的主体区域，这个主体有些情况下可能只是物体的一部分，也可能互相重叠，而分割中我们必须准确的描出属于物体的全部像素。
• 不同的任务下，物体的定义会改变，而我们希望数据的标注尽可能重用。
• 由于现实世界和人类自身的模糊性 (fuzzyness)，以及任务的多样化，经常对于同类物体有不同的标注要求，而我们希望实现数据标注的重用。
• 规则一致性，是数据重用的必要条件。标注规则可以复杂，但只能有一个！
• 规则可以演进，而只要保证统一性，向前向后兼容。当出现多规则，规则间必定是不互通的，显然无法重用。
• 底层的规则尽量是原子的，易组合的，来满足上层的查询需求。

Wordbook

Visual Object Classes，VOC：视觉目标分类
Pattern Analysis, Statistical Modelling and Computational Learning，PASCAL

References

[1] Yongqiang Cheng, https://yongqiang.blog.csdn/
[2] VOC2011 Annotation Guidelines, http://host.robots.ox.ac.uk/pascal/VOC/voc2012/guidelines.html

本文标签： 图像规则指南objectVisual