Data Sets
SMIDS
Sperm Morphology Image Data Set - SMIDS
The SMIDS was collected by using a smartphone based data acquisition approach. Patches were labelled as normal (1021), abnormal (1005) and non-sperm (974) resulting in a three-class data-set. The images were stored in RGB color-space with various sizes. Each image in SMIDS also may include noise, multi-sperm head and mixed tails.
The SMIDS was collected by using a smartphone based data acquisition approach. Patches were labelled as normal (1021), abnormal (1005) and non-sperm (974) resulting in a three-class data-set. The images were stored in RGB color-space with various sizes. Each image in SMIDS also may include noise, multi-sperm head and mixed tails.
SeSViD
Semen Specimen Video Dataset - SeSViD
The SeSViD contains 12 semen specimen videos obtained from different patients (ages from 19 to 37) in the infertility clinic of Istanbul University. Ethical approvals were obtained from the Istanbul University Ethics Committee. Before the video recording, semen samples were stored in a sterile, wide-mouthed, non-toxic 60 ml sample container provided by masturbation from the patients between 9 - 11 am, and were liquefied in a plastic container at approximately 37°C. Then, 5 µL of semen sample was placed in the Makler counting chamber for video recording under 100x magnification using a microscope (Olympus CX31). Videos were recorded at 30 fps in original format. Then, the frame rate was decreased to 5 fps in order to ease the labeling process. Additionally, since the image size is in high quality as 1920 x 1200 resolution in the original format of videos, each frame was resized to 640 x 640 to use in deep learning-based detection model architectures.
The SeSViD contains 12 semen specimen videos obtained from different patients (ages from 19 to 37) in the infertility clinic of Istanbul University. Ethical approvals were obtained from the Istanbul University Ethics Committee. Before the video recording, semen samples were stored in a sterile, wide-mouthed, non-toxic 60 ml sample container provided by masturbation from the patients between 9 - 11 am, and were liquefied in a plastic container at approximately 37°C. Then, 5 µL of semen sample was placed in the Makler counting chamber for video recording under 100x magnification using a microscope (Olympus CX31). Videos were recorded at 30 fps in original format. Then, the frame rate was decreased to 5 fps in order to ease the labeling process. Additionally, since the image size is in high quality as 1920 x 1200 resolution in the original format of videos, each frame was resized to 640 x 640 to use in deep learning-based detection model architectures.
