This is human kinship verification experiment.
The goal of this experiment is to evaluate human ability to recognize family members of various types by viewing pairs of face images, i.e., only faces can be references, as we are looking to compare humans to computer visions algorithms.
If you recognize either of the faces, even if you do not have prior knowledge of blood relationships, please select 'Skip' and move onto the next pair.
The session consists of 200 pairs of face images chosen at random, each grouped according to the pair type. There are 11 types in total: Brother-Brother, Sister-Sister, Siblings (Brother-Sister), Father-Daughter, Father-Son, Mother-Daughter, Mother-Son, Grandfather-Granddaughter, Grandfather-Grandson, Grandmother-Granddaughter, Grandmother-Grandson.
Note that this data is age invariant. In other words, any face can be of any age. For example, a Mother-Daughter pair consisting of a picture of the mother as a child and the daughter as an adult. Regardless of age, verify whether the two faces share the kin relationship (again, listed above the faces/ image window title).
Process is as follows:
Each pair of faces respond with one of the following:
* Click 'Yes' if believed pairs are blood relatives
* Click 'No' if believed pairs do not share kinship
* Click ‘Recognize' if you recognize either of the subjects and or have knowledge of kin/non-kin
Also, a text-field is provided to enter notes for a given current pair (optional).
Any thoughts, reasoning, etc.
Any and all notes are appreciated, but are optional.
Sample Notes may include, but are certainly not limited to, the following:
- Particular feature/ characteristic leading to answer yes/no
- Easy/ more obvious kin/ non-kin pair; what makes it so evident
- Anything that comes to mind when analyzing pair please share.
Lastly, demographic information will allow for a more elaborate analysis of experiment. This information does not need to be provided, though it would be useful and, thus,
Thank you for participating.
Email [email protected] with any questions, issues, or ideas.
Date Created: October 29, 2017
Author: Joseph Robinson and Lisa Gao
The goal of this experiment is to evaluate human ability to recognize family members of various types by viewing pairs of face images, i.e., only faces can be references, as we are looking to compare humans to computer visions algorithms.
If you recognize either of the faces, even if you do not have prior knowledge of blood relationships, please select 'Skip' and move onto the next pair.
The session consists of 200 pairs of face images chosen at random, each grouped according to the pair type. There are 11 types in total: Brother-Brother, Sister-Sister, Siblings (Brother-Sister), Father-Daughter, Father-Son, Mother-Daughter, Mother-Son, Grandfather-Granddaughter, Grandfather-Grandson, Grandmother-Granddaughter, Grandmother-Grandson.
Note that this data is age invariant. In other words, any face can be of any age. For example, a Mother-Daughter pair consisting of a picture of the mother as a child and the daughter as an adult. Regardless of age, verify whether the two faces share the kin relationship (again, listed above the faces/ image window title).
Process is as follows:
Each pair of faces respond with one of the following:
* Click 'Yes' if believed pairs are blood relatives
* Click 'No' if believed pairs do not share kinship
* Click ‘Recognize' if you recognize either of the subjects and or have knowledge of kin/non-kin
Also, a text-field is provided to enter notes for a given current pair (optional).
Any thoughts, reasoning, etc.
Any and all notes are appreciated, but are optional.
Sample Notes may include, but are certainly not limited to, the following:
- Particular feature/ characteristic leading to answer yes/no
- Easy/ more obvious kin/ non-kin pair; what makes it so evident
- Anything that comes to mind when analyzing pair please share.
Lastly, demographic information will allow for a more elaborate analysis of experiment. This information does not need to be provided, though it would be useful and, thus,
Thank you for participating.
Email [email protected] with any questions, issues, or ideas.
Date Created: October 29, 2017
Author: Joseph Robinson and Lisa Gao