携程
Dhurandhar Box Office Day 3: Ranveer Singh's Film Crossed Rs 100 Crore Mark_我的网站
A | Ranveer Singh's Dhurandhar is soaring at the box office, delivering big numbers over the opening weekend. The Aditya Dhar directorial is a spy thriller revolving around a RAW agent and his covert missions in the criminal underworld of Karachi and its link to India. As per industry tracker Sacnilk, Dhurandhar earned Rs 28 crore on its opening day and Rs 32 crore and Rs 43 crore over the weekend, taking the India net collection to Rs 103 crore. The film's Sunday earnings registered an impressive 34.38% jump from its Saturday collection. Overseas, the film earned Rs 21 crore. As per Sacnilk, Dhurandhar had an overall 53.47% Hindi occupancy on Sunday. The film's opening day collections and now the jump over the weekend surpassed predictions made by trade experts, who estimated that it would earn Rs 15-18 crore on day 1. The opening day collection also came very close to Vicky Kaushal's Chhaava's opening day collection of Rs 31 crore, making it a solid comeback for Ranveer Singh. Besides Ranveer, Dhurandhar features an ensemble cast featuring Akshaye Khanna, R Madhavan, Sanjay Dutt, Sara Arjun, Rakesh Bedi and Arjun Rampal among others. Filmfare's Review of Dhurandhar Filmfare rated the film 3.5 out of 5 and said in its review, "At the centre of this world stands Ranveer Singh, playing the hard-edged agent who slips into the shadow world of sleeper operatives as a cross-border conspiracy erupts. It’s a performance of restraint, a departure from his flamboyant screen persona. This Ranveer is raw yet internalised, driven yet fraying. The role demands both physical ferocity and emotional weariness, and he delivers both with remarkable control. His silences speak as loudly as his action beats, grounding the film even when its narrative ambitions begin to overtake its focus." Dhurandhar's sequel, the continuation of the story through a Part 2 feature, will be released in March next year. Also Read: Ranveer Singh’s Dhurandhar soars past Rs 60 Crore, breaks multiple records。 IT之家 7 月 21 日消息，科技媒体 The Decoder 昨日（7 月 20 日）发布博文，报道称谷歌 DeepMind 发布 GenCeption 模型，将预训练的视频生成器重新用于深度估计和分割等经典计算机视觉任务。 IT之家援引博文介绍，大语言模型在学习预测下一个 Token 的时候，在训练过程中往往需要吸收语法、世界知识和上下文关系等内容。 但是在计算机视觉领域，视觉模型缺少等效的训练方法，主要由专业模型主导，包括用于分割的“Segment Anything”和用于深度估计的“Depth Anything”，每个模型都使用其特定的架构。 谷歌 DeepMind 团队为此提出 GenCeption 模型方案，尝试将一个“生成视频”的 AI 模型逆向改造成一个能“理解世界”的视觉分析引擎。 GenCeption 打破了传统计算机视觉“一个任务一个专用模型”的格局，仅凭单一模型就能同时做好深度估计、图像分割、3D 姿态估计、表面法线预测和相机姿态估计等核心视觉任务。 GenCeption 基于阿里巴巴开源视频模型通义万相 Wan2.1 系列训练，与传统扩散模型需多步去噪不同，GenCeption 在一次前向传播中完成预测，从而提升视觉任务处理速度。模型通过文本提示指定任务，可输出深度图、表面法线图、分割掩码，并可处理相机运动表示。
B | 训练数据以合成为主。论文称，数据集仅包含 7500 段视频，由 800 个数字人体模型与 200 段动作捕捉序列组合生成，再通过 Blender 在不同背景和镜头角度下渲染。 泛化方面，GenCeption 几乎只在单人合成视频上训练，但可处理真实多人视频，也可迁移到动物和类人机器人类别。论文称，部分输出细节甚至超过训练时 Blender 渲染结果，可保留猫胡须和单根发丝边缘。 性能方面，论文给出两组处理时间数据：小模型处理 81 帧视频约需 6 秒；大模型参数量为 140 亿，处理同样长度视频约需 10 秒。 参考。
Current article:http://z3bvn.panningbuquanchencaibingzhang.cyou/news/20260826_346.html
Published on:04:22:19
