The latest articles on DEV Community by jongmin kim (@jongmin_kim_26e6d6df88003). https://dev.to/jongmin_kim_26e6d6df88003 https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4233210%2Fb4b8add4-ff10-407b-80a5-853954ebff63.png https://dev.to/jongmin_kim_26e6d6df88003 en jongmin kim Wed, 11 Nov 2026 09:00:00 +0000 https://dev.to/jongmin_kim_26e6d6df88003/what-i-measured-when-searching-30-minutes-of-dashcam-footage-1jfo https://dev.to/jongmin_kim_26e6d6df88003/what-i-measured-when-searching-30-minutes-of-dashcam-footage-1jfo <p>A single search time can hide where a video review tool actually makes you wait. While testing Dashcam Finder, a Windows app I am building at JM Soft, I timed preparation, analysis, and the first set of suggestions separately.</p> <h2> The test </h2> <p>I used a 30-minute test file made from 29 minutes and 30 seconds of public road footage and a 30-second public dashcam excerpt. I sampled one frame per second and searched for <code>a traffic sign</code>.</p> <div class="table-wrapper-paragraph"><table> <thead> <tr> <th>Stage</th> <th>Time in this run</th> </tr> </thead> <tbody> <tr> <td>Startup preparation</td> <td>44.14 s</td> </tr> <tr> <td>Video analysis</td> <td>123.75 s</td> </tr> <tr> <td>Initial search after analysis</td> <td>3.32 s</td> </tr> <tr> <td>Combined time through first suggestions</td> <td>171.21 s</td> </tr> </tbody> </table></div> <p>These numbers come from one run. The initial search took 3.32 seconds because the video had already been analyzed. A separate detailed search added 111.62 seconds in this test.</p> <h2> How I checked the result </h2> <p>I selected a suggested timestamp and opened the original footage to check what was actually visible. A suggestion gives me a place to look. The original video is what I use to confirm the result. Sampling once per second can miss a detail that appears only between samples.</p> <p>The same distinction matters for a small car in the distance. The app can suggest vehicle models to compare, and the reviewer can inspect the original scene before drawing a conclusion.</p> <p><a href="https://today-tip.com/video-search-preview/#traffic-sign-demo" rel="noopener noreferrer">Watch the traffic sign demo</a> or <a href="https://today-tip.com/video-search-preview/#vehicle-model-demo" rel="noopener noreferrer">see the separate small-vehicle comparison</a>.</p> <p>If you have built a video review workflow, which would you time separately next: seeking to the original frame, or narrowing the candidate list?</p> <p><em>Disclosure: I build Dashcam Finder. This article was drafted with AI assistance using my test records.</em></p> buildinpublic performance software