Ç¥ÁØÈ­ Âü¿©¾È³»

TTAÀÇ Ç¥ÁØÇöȲ

Ȩ > Ç¥ÁØÈ­ °³¿ä > TTAÀÇ Ç¥ÁØÇöȲ

Ç¥ÁعøÈ£ TTAK.KO-11.0310 ±¸Ç¥ÁعøÈ£
Á¦°³Á¤ÀÏ 2022-12-07 ÃÑÆäÀÌÁö 26
ÇѱÛÇ¥Áظí JSON ÇüÅÂÀÇ ÀÚÀ²ÁÖÇà ÇнÀ¿ë µ¥ÀÌÅÍ Ç°Áú °ËÁõ Áöħ – Á¤È®¼º ¹× ´Ù¾ç¼º
¿µ¹®Ç¥Áظí Guidelines for Quality Verification of Training Data of Autonomous Driving in JSON Format – Accuracy and Diversity
Çѱ۳»¿ë¿ä¾à 4Â÷ »ê¾÷Çõ¸í ±â¼úÀÇ Çٽɱâ¼ú Áß ÇϳªÀÎ ÀÚÀ²ÁÖÇàÀº ´Ü¼ø À̵¿ ¼ö´Ü °³³äÀ» ³Ñ¾î¼­ Á¢±Ù¼º, ¾ÈÀü¼º, Æí¸®¼º Ãø¸é¿¡¼­ Çõ½ÅÀûÀÎ º¯È­¸¦ °¡Á®¿À¸é¼­ ÀÚµ¿Â÷ »ê¾÷ Àü¹ÝÀÇ Æз¯´ÙÀÓÀ» À̲ø¾î³»°í ÀÖ´Ù. ÀÌ·¯ÇÑ Æз¯´ÙÀÓ Àüȯ¿¡¼­ ÀÚµ¿Â÷ »ê¾÷À» ³Ñ¾î ¸ðºô¸®Æ¼ »ê¾÷À» ¼±µµÇϱâ À§Çؼ­´Â ÀÚÀ²ÁÖÇàÀÇ ±â¼úÀû¡¤¾ÈÀü¼º Ãø¸éÀÇ ½Å·Ú¼º È®º¸°¡ ¹«¾ùº¸´Ù Áß¿äÇÏ´Ù.

ÀÚÀ²ÁÖÇà ½Å·Ú¼º È®º¸¸¦ À§ÇØ ÀÚÀ²ÁÖÇà ÇÙ½É ±â¼ú Áß °¡Àå Áß¿äÇÑ ±â¼ú ¿ä¼ÒÀÎ ÀΰøÁö´É ¼ÒÇÁÆ®¿þ¾îÀÇ °íµµÈ­°¡ ÇʼöÀûÀÌ´Ù. ÀΰøÁö´É ¼ÒÇÁÆ®¿þ¾î´Â ÀÚµ¿Â÷ÀÇ ºê·¹ÀÎ ¿ªÇÒ·Î °¢Á¾ ¼¾¼­¸¦ ÅëÇØ ¼öÁýµÈ ÁÖÇàȯ°æ µ¥ÀÌÅ͸¦ Åä´ë·Î ÁÖÇàȯ°æ, À§Ä¡ ÀÎ½Ä ¹× ¸ÊÇÎ, ÀÎÁö, ÆÇ´Ü, Á¦¾î¸¦ ÅëÇØ ÁÖº¯ ±³Åëȯ°æ Á¤º¸¸¦ ÆľÇÇÏ¿© ÀÚÀ²ÁÖÇàÀ» °¡´ÉÇÏ°Ô ÇÑ´Ù.

ÀΰøÁö´É ¼ÒÇÁÆ®¿þ¾î °íµµÈ­´Â ´ë±Ô¸ð·Î ¼öÁýµÈ µ¥ÀÌÅ͸¦ ÇнÀÇÔÀ¸·Î½á ÁøÀüµÇ¸ç, °íµµÈ­ °úÁ¤Àº ÇнÀ µ¥ÀÌÅÍÀÇ Ç°Áú¿¡ µû¶ó ÀΰøÁö´É ¼ÒÇÁÆ®¿þ¾îÀÇ ¼º´ÉÀÌ Á¿ìµÈ´Ù. ÇнÀ¿ë µ¥ÀÌÅ͸¦ »ç¿ëÇÏ´Â ÀÚÀ²ÁÖÇà ÀΰøÁö´É ¼ÒÇÁÆ®¿þ¾îÀÇ °æ¿ì, µ¥ÀÌÅ͸¦ È°¿ëÇϱâ Àü¿¡ µ¥ÀÌÅÍ¿¡ ´ëÇÑ Ç°ÁúÆò°¡¡¤°ËÁõÀ» °í·ÁÇØ¾ß ÇÑ´Ù.

ÃÖ±Ù Á¤ºÎ¿¡¼­ õ¹®ÇÐÀûÀÎ ºñ¿ëÀ» µé¿© ±¸ÃàÇÑ ¡®µ¥ÀÌÅÍ ´ï¡¯ÀÇ ÀÚÀ²ÁÖÇà µ¥ÀÌÅʹ ǥÁØ ÁöħÀÌ ¸¶·ÃµÇÁö ¾Ê¾Æ ±¸Ãà ±â°ü¸¶´Ù µ¥ÀÌÅÍÀÇ ±¸Á¶°¡ Á¦°¢°¢À̾ È°¿ë Ãø¸é¿¡¼­ ¾î·Á¿òÀÌ Á¸ÀçÇÑ´Ù. ¶ÇÇÑ ÀϺΠÆíÇâµÈ µ¥ÀÌÅÍ·Î ÀÎÇØ ÀΰøÁö´É ¼º´É ¹®Á¦¸¦ ³Ñ¾î¼­ À±¸®Àû¡¤»çȸÀû ¹®Á¦¸¦ ¹ß»ý½Ãų ¼ö ÀÖ´Ù.

µû¶ó¼­ º» Ç¥ÁØ¿¡¼­´Â ±¸ÃàµÈ ÀÚÀ²ÁÖÇà ÇнÀ¿ë µ¥ÀÌÅÍ È°¿ëÀ» À§ÇØ ¡®±¸¹®Àû Á¤È®¼º¡¯°ú ¡®Åë°èÀû ÃæºÐ¼º¡¤±ÕÀϼº¡¯À¸·Î µÎ °¡Áö ¹üÁÖÀÇ Á¤·®Àû Ç°ÁúÁöÇ¥¸¦ Á¦½ÃÇϸç, ÀΰøÁö´É ¼ÒÇÁÆ®¿þ¾î°¡ ÀÚÀ²ÁÖÇà ÇнÀ¿ë µ¥ÀÌÅ͸¦ È°¿ëÇϱâ À§ÇØ °øÅëÀûÀ¸·Î °®Ãß¾î¾ßÇÒ ±¸Á¶ Á¤È®¼º ¹× ´Ù¾ç¼ºÀ» ¹Ý¿µÇÑ ½ºÅ°¸¶ ±¸Á¶¸¦ Á¦½ÃÇϴµ¥ ¸ñÀûÀÌ ÀÖ´Ù.
¿µ¹®³»¿ë¿ä¾à Autonomous driving, one of the core technologies of the 4th industrial revolution technology, is leading the paradigm of the entire automobile industry by bringing innovative changes in accessibility, safety, and convenience beyond the concept of a simple means of transportation. In order to lead the mobility industry beyond the automobile industry in this paradigm shift, securing reliability in terms of technology and safety of autonomous driving is of utmost importance.

In order to secure the reliability of autonomous driving, the advancement of artificial intelligence software, which is the most important technology element among core autonomous driving technologies, is essential. The artificial intelligence software serves as the brain of the car and enables autonomous driving by grasping the surrounding traffic environment information through driving environment, location recognition and mapping, recognition, judgment, and control based on driving environment data collected through various sensors.

The advancement of artificial intelligence software progresses by training data collected on a large scale, and the performance of artificial intelligence software depends on the quality of the training data in the advancement process. In the case of autonomous driving artificial intelligence software that uses data for training, quality evaluation and verification of data should be considered before using data.

The autonomous driving data of the ¡®Data Dam¡¯, which the government recently built at an astronomical cost, does not have standard guidelines, so the data structure is different for each building institution, so there are difficulties in using it. In addition, some biased data can cause ethical and social issues beyond AI performance issues.

Therefore, in this standard, two categories of quantitative quality indicators are presented as 'syntactic accuracy' and 'statistical sufficiency¡¤uniformity¡¯ to utilize the constructed training data for autonomous driving. It aims to present a schema structure that reflects the structural accuracy and diversity that must be in common for this purpose.
±¹Á¦Ç¥ÁØ
°ü·ÃÆÄÀÏ TTAK.KO-11.0310.pdf TTAK.KO-11.0310.pdf            

ÀÌÀü
i-PIN ¼­ºñ½º Àü´Þ ¸Þ½ÃÁö Çü½Ä
´ÙÀ½
ÄÄÇ»ÅÍ Æ÷·»½ÄÀ» À§ÇÑ µðÁöÅÐ µ¥ÀÌÅÍ ¼öÁýµµ±¸ ¿ä±¸»çÇ×