Class aware object marking tool

Abstract

Systems and methods for class aware object marking are provided. For example, an indication of a selected object type of a plurality of alternative object types may be received. Further, in some examples, an image may be presented to a user. Further, in some examples, an input to an object selection tool may be received from the user. Further, in some examples, in response to a first selected object type and the input to the object selection tool, a first region of the image corresponding to an object of the first selected object type in the image may be selected, and in response to a second selected object type and the input to the object selection tool, a second region of the image corresponding to an object of the second selected object type in the image may be selected.

1 . A method for class aware object marking, the method comprising:
 receiving an indication of a selected object type of a plurality of alternative object types;   presenting an image to a user;   receiving from the user an input to an object selection tool;   in response to a first selected object type and the input to the object selection tool, selecting a first region of the image corresponding to an object of the first selected object type in the image; and   in response to a second selected object type and the input to the object selection tool, selecting a second region of the image corresponding to an object of the second selected object type in the image, where the first region differs from the second region.   
     
     
         2 . The method of  claim 1 , further comprising:
 presenting to the user the plurality of alternative object types; and   receiving from the user the indication of the selected object type of the plurality of alternative object types.   
     
     
         3 . The method of  claim 1 , further comprising analyzing the image to select the selected object type of the plurality of alternative object types. 
     
     
         4 . The method of  claim 1 , wherein the input to the object selection tool comprises one or more points in the image, and the selected region of the image includes the one or more points and at least one additional pixel of the image. 
     
     
         5 . The method of  claim 4 , wherein the input to the object selection tool comprises one or more additional points in the image, and the selected region of the image do not include any of the one or more additional points. 
     
     
         6 . The method of  claim 1 , wherein the input to the object selection tool comprises one or more scribbles in the image, and the selected region of the image includes the one or more scribbles and at least one additional pixel of the image. 
     
     
         7 . The method of  claim 6 , wherein the input to the object selection tool comprises one or more additional scribbles in the image, and the selected region of the image do not include any of the one or more additional scribbles. 
     
     
         8 . The method of  claim 1 , wherein the input to the object selection tool comprises a bounding polygon, and the selected region of the image is included within the bounding polygon. 
     
     
         9 . The method of  claim 1 , wherein the first selected object type is person together objects held by the person and the second selected object type is person not including objects held by the person. 
     
     
         10 . The method of  claim 1 , wherein the first selected object type is person including occluded parts and the second selected object type is person except for occluded parts. 
     
     
         11 . The method of  claim 1 , wherein the first selected object type is male and the second selected object type is female. 
     
     
         12 . The method of  claim 1 , wherein the first selected object type is face including hair and the second selected object type is face except for hair. 
     
     
         13 . The method of  claim 1 , wherein the first selected object type is a single instance of an object of a specified object type and the second selected object type is one or more instances of objects of the specified object type. 
     
     
         14 . The method of  claim 1 , wherein the first selected object type is a person and the second selected object type is a vehicle. 
     
     
         15 . The method of  claim 1 , wherein the first selected object type is a car and the second selected object type is a truck. 
     
     
         16 . The method of  claim 1 , further comprising:
 after the selection of the first region of the image corresponding to the object of the first selected object type in the image, receiving from the user an indication of a desire to change the selected object type to an updated object type, the updated object type differs from the first selected object type; and   in response to the received indication of the desire to change the selected object type, using the updated object type and the input to the object selection tool to select a modified region of the image corresponding to an object of the updated object type in the image.   
     
     
         17 . The method of  claim 1 , further comprising updating a dataset of images and masks according to the selected region of the image. 
     
     
         18 . The method of  claim 1 , further comprising:
 analyzing the image and the input to the object selection tool to determine whether a region of the image corresponding to the input to the object selection tool depicts an object of the selected object type; and   in response to a determination that the region of the image corresponding to the input to the object selection tool does not depict an object of the selected object type, provide an indication to the user.   
     
     
         19 . A system for class aware object marking, the system comprising:
 at least one processor configured to:
 receive an indication of a selected object type of a plurality of alternative object types; 
 present an image to a user; 
 receive from the user an input to an object selection tool; 
 in response to a first selected object type and the input to the object selection tool, select a first region of the image corresponding to an object of the first selected object type in the image; and 
 in response to a second selected object type and the input to the object selection tool, select a second region of the image corresponding to an object of the second selected object type in the image, where the first region differs from the second region. 
   
     
     
         20 . A non-transitory computer readable medium storing data and computer implementable instructions for carrying out a method for class aware object marking, the method comprising:
 receiving an indication of a selected object type of a plurality of alternative object types;   presenting an image to a user;   receiving from the user an input to an object selection tool;   in response to a first selected object type and the input to the object selection tool, selecting a first region of the image corresponding to an object of the first selected object type in the image; and   in response to a second selected object type and the input to the object selection tool, selecting a second region of the image corresponding to an object of the second selected object type in the image, where the first region differs from the second region.