Meta SAM3 Celtics Knicks tracking

by Gifamoss: Basketball

Meta has released SAM3 at no cost as fully open source material and the performance stands out for video based analysis work. Select any visible item inside footage including an athlete a ball a shadow or even one figure within a larger group and the system maintains consistent identification from start to finish across successive frames. Observe the provided example from a contest between the Celtics and the Knicks. Ten athletes appear amid quick transitions overlapping movement and brief jersey visibility yet SAM3 assigns each participant their correct name while generating an immediate two dimensional layout of all locations. Organizations in professional basketball have historically paid substantial yearly fees to firms including Second Spectrum and Stats Perform for comparable detailed reviews. The cost comparison proves striking. Annual NBA tracking agreements fall between one million and five million dollars per team. Equivalent results achieved through SAM3 require only a weekend of effort with no financial outlay and demand nothing beyond a single initial selection. The method relies on promptable segmentation. Supplying a descriptive term a point or a bounding area allows the model to recognize the chosen subject throughout the unbroken video sequence treating the element as one continuous entity instead of separate images. Additional opportunities emerge for numerous other organizations. College athletic departments high schools soccer development programs mixed martial arts training sites and esports teams all seek advanced performance metrics without access to large budgets. Independent reviewers now provide full game analyses priced from five hundred to two thousand dollars each. A single individual working with one computer over a short period can therefore create an ongoing service offering.