Labeling with<\/h2><\/div><\/div><\/div>higher speed and accuracy<\/h2><\/div><\/div><\/div><\/div>[vc_column_text]<\/p>\r\n\ud83d\udd11<\/strong> In 5 minutes you will learn:
\r\n<\/strong><\/span><\/p>\r\n <\/p>\r\n
\r\n- Labeling problems Datumo solves<\/li>\r\n
- Effective data acquisition method via semi-automated labeling<\/li>\r\n<\/ul>\r\n
[\/vc_column_text][vc_raw_html]JTNDbWV0YSUyMGh0dHAtZXF1aXYlM0QlMjJyZWZyZXNoJTIyJTIwY29udGVudCUzRCUyMjAlM0IlMjB1cmwlM0RodHRwcyUzQSUyRiUyRmRhdHVtby5jb20lMkZlbiUyRmRhdHVtb3Mtc2VtaS1hdXRvbWF0ZWQtbGFiZWxpbmctdGVjaG5vbG9neS10by1zYXZlLXRpbWUtYW5kLWNvc3QlMkYlMjIlM0U=[\/vc_raw_html]<\/div><\/div>[\/vc_column][\/vc_row][vc_row pix_particles_check=””][vc_column]
<\/div>[vc_column_text]Q. Label the cat from the image below.<\/strong><\/span><\/p>\r\n <\/p>\r\n
![\"\"](\"https:\/\/blog.datumo.com\/en\/wp-content\/uploads\/2023\/03\/\ubc18\uc790\ub3d9\ub808\uc774\ube14\ub9c1-1536x1021-1-1024x681.webp\")
<\/p>\r\n
<\/p>\r\n
\u201cHow can we label an object that is occluded by another object?\u201d<\/em><\/span><\/p>\r\n <\/p>\r\n
Humans can distinguish the cat from the tree. However, an AI model trained with general cat images may automatically label the whole tree and the cat as a \u201ccat\u201d since it does not acknowledge the occlusion. This could be a problem since part of the cat is simply hidden behind the tree, not a separate object. On the other hand, if manual segmentation is carried out, the accuracy may go up but as more control points are made, more time and cost will be spent.<\/span><\/p>\r\nIn order to solve this problem, Datumo came up with the <Semi-automated Labeling System> to reduce the cost and time while increasing the accuracy.<\/span>[\/vc_column_text]<\/div><\/div><\/div><\/div>[\/vc_column][\/vc_row][vc_section full_width=”stretch_row” pix_over_visibility=”” css=”.vc_custom_1678860341992{padding-top: 80px !important;padding-bottom: 80px !important;background-color: #ffffff !important;}” el_id=”pix_section_program”][vc_row full_width=”stretch_row” pix_particles_check=””][vc_column content_align=”text-center” offset=”vc_col-lg-offset-0 vc_col-lg-12 vc_col-md-offset-1 vc_col-md-10″]Semi-automated Labeling<\/h2><\/div>[vc_column_text css=”.vc_custom_1678860258573{border-bottom-width: 1pt !important;padding-bottom: 120px !important;border-bottom-color: rgba(0,0,0,0.2) !important;border-bottom-style: solid !important;}”]<\/p>\r\n
Semi-automated labeling is a technology of enabling smoother segmenation process<\/strong> via AI technology.<\/p>\r\n![\"\"](\"https:\/\/blog.datumo.com\/en\/wp-content\/uploads\/2023\/03\/\uc601\ubb381.webp\")
<\/p>\r\n
<\/p>\r\n
The annotator can simply make necessary modifications on the image that has been pro-processed by AI.<\/p>\r\n
[\/vc_column_text][\/vc_column][\/vc_row][\/vc_section][vc_row pix_particles_check=””][vc_column]
\u2714\ufe0f Time spent in labeling is reduced by<\/span><\/br> almost 50%, compared to manual labeling<\/span><\/h4><\/div><\/div>\u2714\ufe0f Semi-automation enabled,<\/span>regardless of domain<\/span>, <\/span>with minimum of 400<\/span> manual annotation<\/span><\/h4><\/div><\/div><\/div>[vc_row_inner][vc_column_inner width=”1\/2″]mIoU:<\/span>0<\/span>%<\/span><\/h1><\/div>Simple 400 cow dataset<\/h6><\/div><\/div>\n[\/vc_column_inner][vc_column_inner width=”1\/2″]mIoU:<\/span>0<\/span>%<\/span><\/h1><\/div>Total combination of 400 cityscape dataset<\/h6><\/div><\/div>\n[\/vc_column_inner][\/vc_row_inner]<\/div>[\/vc_column][\/vc_row][vc_section full_width=”stretch_row” pix_over_visibility=”” css=”.vc_custom_1679035464026{padding-top: 80px !important;padding-bottom: 80px !important;background-color: #ffffff !important;}”][vc_row full_width=”stretch_row” pix_particles_check=””][vc_column content_align=”text-center” offset=”vc_col-lg-offset-0 vc_col-lg-12 vc_col-md-offset-1 vc_col-md-10″][vc_video link=”https:\/\/youtu.be\/qNX-yori-9g” align=”center”][vc_column_text css=”.vc_custom_1679035612984{padding-top: 60px !important;padding-bottom: px !important;}”]<\/p>\r\nUp to 40~50% of time<\/strong> spent in labeling can be reduced by such semi-automated labeling. Datumo\u2019s technology enables semi-automated labeling for any data, regardless of its domain, with minimum of 400 labeled data<\/strong>. This makes the technology universal.<\/p>\r\n[\/vc_column_text]
<\/div>
higher speed and accuracy<\/h2><\/div><\/div><\/div><\/div>[vc_column_text]<\/p>\r\n\ud83d\udd11<\/strong> In 5 minutes you will learn:
\r\n<\/strong><\/span><\/p>\r\n <\/p>\r\n
\r\n- Labeling problems Datumo solves<\/li>\r\n
- Effective data acquisition method via semi-automated labeling<\/li>\r\n<\/ul>\r\n
[\/vc_column_text][vc_raw_html]JTNDbWV0YSUyMGh0dHAtZXF1aXYlM0QlMjJyZWZyZXNoJTIyJTIwY29udGVudCUzRCUyMjAlM0IlMjB1cmwlM0RodHRwcyUzQSUyRiUyRmRhdHVtby5jb20lMkZlbiUyRmRhdHVtb3Mtc2VtaS1hdXRvbWF0ZWQtbGFiZWxpbmctdGVjaG5vbG9neS10by1zYXZlLXRpbWUtYW5kLWNvc3QlMkYlMjIlM0U=[\/vc_raw_html]<\/div><\/div>[\/vc_column][\/vc_row][vc_row pix_particles_check=””][vc_column]
<\/div>[vc_column_text]Q. Label the cat from the image below.<\/strong><\/span><\/p>\r\n <\/p>\r\n
<\/p>\r\n
<\/p>\r\n
\u201cHow can we label an object that is occluded by another object?\u201d<\/em><\/span><\/p>\r\n <\/p>\r\n
Humans can distinguish the cat from the tree. However, an AI model trained with general cat images may automatically label the whole tree and the cat as a \u201ccat\u201d since it does not acknowledge the occlusion. This could be a problem since part of the cat is simply hidden behind the tree, not a separate object. On the other hand, if manual segmentation is carried out, the accuracy may go up but as more control points are made, more time and cost will be spent.<\/span><\/p>\r\nIn order to solve this problem, Datumo came up with the <Semi-automated Labeling System> to reduce the cost and time while increasing the accuracy.<\/span>[\/vc_column_text]<\/div><\/div><\/div><\/div>[\/vc_column][\/vc_row][vc_section full_width=”stretch_row” pix_over_visibility=”” css=”.vc_custom_1678860341992{padding-top: 80px !important;padding-bottom: 80px !important;background-color: #ffffff !important;}” el_id=”pix_section_program”][vc_row full_width=”stretch_row” pix_particles_check=””][vc_column content_align=”text-center” offset=”vc_col-lg-offset-0 vc_col-lg-12 vc_col-md-offset-1 vc_col-md-10″]Semi-automated Labeling<\/h2><\/div>[vc_column_text css=”.vc_custom_1678860258573{border-bottom-width: 1pt !important;padding-bottom: 120px !important;border-bottom-color: rgba(0,0,0,0.2) !important;border-bottom-style: solid !important;}”]<\/p>\r\n
Semi-automated labeling is a technology of enabling smoother segmenation process<\/strong> via AI technology.<\/p>\r\n<\/p>\r\n
<\/p>\r\n
The annotator can simply make necessary modifications on the image that has been pro-processed by AI.<\/p>\r\n
[\/vc_column_text][\/vc_column][\/vc_row][\/vc_section][vc_row pix_particles_check=””][vc_column]
\u2714\ufe0f Time spent in labeling is reduced by<\/span><\/br> almost 50%, compared to manual labeling<\/span><\/h4><\/div><\/div>\u2714\ufe0f Semi-automation enabled,<\/span>regardless of domain<\/span>, <\/span>with minimum of 400<\/span> manual annotation<\/span><\/h4><\/div><\/div><\/div>[vc_row_inner][vc_column_inner width=”1\/2″]mIoU:<\/span>0<\/span>%<\/span><\/h1><\/div>Simple 400 cow dataset<\/h6><\/div><\/div>\n[\/vc_column_inner][vc_column_inner width=”1\/2″]mIoU:<\/span>0<\/span>%<\/span><\/h1><\/div>Total combination of 400 cityscape dataset<\/h6><\/div><\/div>\n[\/vc_column_inner][\/vc_row_inner]<\/div>[\/vc_column][\/vc_row][vc_section full_width=”stretch_row” pix_over_visibility=”” css=”.vc_custom_1679035464026{padding-top: 80px !important;padding-bottom: 80px !important;background-color: #ffffff !important;}”][vc_row full_width=”stretch_row” pix_particles_check=””][vc_column content_align=”text-center” offset=”vc_col-lg-offset-0 vc_col-lg-12 vc_col-md-offset-1 vc_col-md-10″][vc_video link=”https:\/\/youtu.be\/qNX-yori-9g” align=”center”][vc_column_text css=”.vc_custom_1679035612984{padding-top: 60px !important;padding-bottom: px !important;}”]<\/p>\r\nUp to 40~50% of time<\/strong> spent in labeling can be reduced by such semi-automated labeling. Datumo\u2019s technology enables semi-automated labeling for any data, regardless of its domain, with minimum of 400 labeled data<\/strong>. This makes the technology universal.<\/p>\r\n[\/vc_column_text]
<\/div>
\ud83d\udd11<\/strong> In 5 minutes you will learn: <\/p>\r\n [\/vc_column_text][vc_raw_html]JTNDbWV0YSUyMGh0dHAtZXF1aXYlM0QlMjJyZWZyZXNoJTIyJTIwY29udGVudCUzRCUyMjAlM0IlMjB1cmwlM0RodHRwcyUzQSUyRiUyRmRhdHVtby5jb20lMkZlbiUyRmRhdHVtb3Mtc2VtaS1hdXRvbWF0ZWQtbGFiZWxpbmctdGVjaG5vbG9neS10by1zYXZlLXRpbWUtYW5kLWNvc3QlMkYlMjIlM0U=[\/vc_raw_html]<\/div><\/div>[\/vc_column][\/vc_row][vc_row pix_particles_check=””][vc_column] <\/p>\r\n <\/p>\r\n \u201cHow can we label an object that is occluded by another object?\u201d<\/em><\/span><\/p>\r\n <\/p>\r\n Humans can distinguish the cat from the tree. However, an AI model trained with general cat images may automatically label the whole tree and the cat as a \u201ccat\u201d since it does not acknowledge the occlusion. This could be a problem since part of the cat is simply hidden behind the tree, not a separate object. On the other hand, if manual segmentation is carried out, the accuracy may go up but as more control points are made, more time and cost will be spent.<\/span><\/p>\r\n In order to solve this problem, Datumo came up with the <Semi-automated Labeling System> to reduce the cost and time while increasing the accuracy.<\/span>[\/vc_column_text]<\/div><\/div> Semi-automated labeling is a technology of enabling smoother segmenation process<\/strong> via AI technology.<\/p>\r\n <\/p>\r\n The annotator can simply make necessary modifications on the image that has been pro-processed by AI.<\/p>\r\n [\/vc_column_text][\/vc_column][\/vc_row][\/vc_section][vc_row pix_particles_check=””][vc_column] Up to 40~50% of time<\/strong> spent in labeling can be reduced by such semi-automated labeling. Datumo\u2019s technology enables semi-automated labeling for any data, regardless of its domain, with minimum of 400 labeled data<\/strong>. This makes the technology universal.<\/p>\r\n [\/vc_column_text]
\r\n<\/strong><\/span><\/p>\r\n\r\n
<\/p>\r\nSemi-automated Labeling<\/h2><\/div>[vc_column_text css=”.vc_custom_1678860258573{border-bottom-width: 1pt !important;padding-bottom: 120px !important;border-bottom-color: rgba(0,0,0,0.2) !important;border-bottom-style: solid !important;}”]<\/p>\r\n
<\/p>\r\n\u2714\ufe0f Time spent in labeling is reduced by<\/span><\/br> almost 50%, compared to manual labeling<\/span><\/h4><\/div>
\u2714\ufe0f Semi-automation enabled,<\/span>regardless of domain<\/span>, <\/span>with minimum of 400<\/span> manual annotation<\/span><\/h4><\/div>
mIoU:<\/span>0<\/span>%<\/span><\/h1><\/div>
Simple 400 cow dataset<\/h6><\/div><\/div>\n[\/vc_column_inner][vc_column_inner width=”1\/2″]
mIoU:<\/span>0<\/span>%<\/span><\/h1><\/div>
Total combination of 400 cityscape dataset<\/h6><\/div><\/div>\n[\/vc_column_inner][\/vc_row_inner]