OpenStreetMap is an universal, openly available, crowdsourced and free geographic database covering global street networks. Researchers utilize OSM for a diverse range of bicycle research topics, including route choice, bikeability assessment and simulation. However, the crowdsourced nature of OSM does reveal substantial gaps such as missing, mistagged cycling paths or land use attributes in certain areas. To mitigate this problem, georeferenced Street View Imagery (SVI) and computer vision (CV) tasks can be performed for feature extraction. Similar approaches have been utilized for bikeability (Ito & Biljecki 2021), walkability (Nagata et al. 2020) and bicycle infrastructure classification issues (Saxton 2022). These studies, however, do not focus on OSM data enrichment. This contribution presents a CV method, a fine-tuned Mask2Former approach (Cheng et al. 2022) that is deployed and used for inference on the SVI. It incorporates a Detection Transformer (Carion et al. 2020) fine-tuned on German traffic signs for further analysis. The features extracted (e.g. proportion of roads, buildings, greenery, visible sky) are matched to the OSM network and attributed to their respective network edges, which can be utilized for further research goals, such as route choice modeling. For the proof of concept, around 25.000 SVI were collected in the city center of Bietigheim-Bissingen, Germany, covering an area of around 4,1 km². Features were extracted using aforementioned CV methods, aggregated and attributed to the OSM network using valhalla. The developed method helps closing knowledge and data gaps and explore and identify novel attributes to enrich bicycle research. The next steps include a thorough adaptation of the CV tasks to German cycling infrastructure as well as validation between cycling infrastructure present in OSM and SVI. Prospectively, CV-enriched network data will be able to improve bicycle research tasks in general.