Large-scale video-language pre-training has made remarkable strides in advancing video-language understanding tasks. However, the heavy computational burden of video encoding remains a formidable efficiency bottleneck, particularly for long-form videos. These videos contain massive visual tokens due to their inherent 3D properties and spatiotemporal redundancy, making it challenging to capture complex temporal and spatial relationships. To tackle this issue, we propose an efficient method called TEmporal-Spatial Token Aggregation (TESTA). TESTA condenses video semantics by adaptively aggregating similar frames, as well as similar patches within each frame. TESTA can reduce the number of visual tokens by 75% and thus accelerate video encoding. Building upon TESTA, we introduce a pre-trained video-language model equipped with a divided space-time token aggregation module in each video encoder block. We evaluate our model on five datasets for paragraph-to-video retrieval and long-form VideoQA tasks. Experimental results show that TESTA improves computing efficiency by 1.7 times, and achieves significant performance gains from its scalability in processing longer input frames, e.g., +13.7 R@1 on QuerYD and +6.5 R@1 on Condensed Movie.

I have an alarm set for 3 a.m. on Sunday morning, which is the earliest that we can start making restaurant reservations for our trip. There's even a spreadsheet with all the restaurants we want to visit, as meticulously researched from videos and handy guides from places like Disney Food Blog and WDW Prep School. So I was more than receptive when someone recommended that I check out TouringPlans.com It's a premium service, with subscriptions priced starting at $16 per year. It has seven full-time employees on the staff, and 12 part-time employees, says Testa, as well as a business in publishing "Unofficial Guides" to Disney World, Disneyland, and other popular tourist destinations.

I have an alarm set for 3 a.m. on Sunday morning, which is the earliest that we can start making restaurant reservations for our trip. There's even a spreadsheet with all the restaurants we want to visit, as meticulously researched from videos and handy guides from places like Disney Food Blog and WDW Prep School. So I was more than receptive when someone recommended that I check out TouringPlans.com-- a site that uses complex algorithms to help you plan the perfect vacation at Disney World, Disneyland, or a handful of other theme parks like Universal Studios Orlando. It's a premium service, with subscriptions priced starting at $16 per year. It has seven full-time employees on the staff, and 12 part-time employees, says Testa, as well as a business in publishing "Unofficial Guides" to Disney World, Disneyland, and other popular tourist destinations.

He scrutinized maps and a detailed timetable. He even deployed a secret weapon: artificial-intelligence research to chart a course through death-defying drops, torrents of water and fiery heat. And when this adventurer clambered out of a floating log last year, he had reached his holy grail: visiting -- in a single day -- each of the 41 operating rides, attractions and shows at the Everest of theme parks, Walt Disney World's Magic Kingdom. "It was a challenge, and I loved the idea," said Vosburgh, 45, an information technology manager for Dell Computer Co. in Austin, Texas. "There are a lot of things at Disney World that I had not done. Why would I go on teacups when there are mountains to ride? This way I was forced to try everything."