File tree Expand file tree Collapse file tree
Expand file tree Collapse file tree Original file line number Diff line number Diff line change @@ -90,7 +90,7 @@ <h1 id="paper-title"><span class="hero-project">EgoHumanoid-V2:</span> <span cla
9090< section aria-labelledby ="visual-title " class ="section " id ="visual-alignment "> < div class ="wrap ">
9191< div class ="section-heading "> < h2 id ="visual-title "> Visual Alignment</ h2 > </ div >
9292
93- < div class ="task-matrix-row visual-workflow " data-matrix-row ="visual-laptop " data-sync ="true "> < div class ="matrix-heading "> < h3 > Close Laptop</ h3 > < button aria-pressed ="false " class ="button small matrix-play " type ="button "> Play workflow</ button > </ div > < div class ="task-matrix-grid "> < figure class ="media-card matrix-card "> < div class ="video-frame poster-shell "> < img alt ="" aria-hidden ="true " class ="video-thumbnail " decoding ="async " loading ="lazy " src ="assets/images/visual-laptop-1.jpg "/> < video aria-label ="Close Laptop: Human RGB " class ="poster-pending " controls ="" data-src ="assets/videos/visual-laptop-1.mp4 " loop ="" muted ="" playsinline ="" poster ="assets/images/visual-laptop-1.jpg " preload ="none "> </ video > </ div > < figcaption > < span class ="workflow-number "> 01</ span > Human RGB</ figcaption > </ figure > < figure class ="media-card matrix-card "> < div class ="video-frame poster-shell "> < img alt ="" aria-hidden ="true " class ="video-thumbnail " decoding ="async " loading ="lazy " src ="assets/images/visual-laptop-2.jpg "/> < video aria-label ="Close Laptop: Arm-removal mask " class ="poster-pending " controls ="" data-src ="assets/videos/visual-laptop-2.mp4 " loop ="" muted ="" playsinline ="" poster ="assets/images/visual-laptop-2.jpg " preload ="none "> </ video > </ div > < figcaption > < span class ="workflow-number "> 02</ span > Arm-removal mask</ figcaption > </ figure > < figure class ="media-card matrix-card "> < div class ="video-frame poster-shell "> < img alt ="" aria-hidden ="true " class ="video-thumbnail " decoding ="async " loading ="lazy " src ="assets/images/visual-laptop-3.jpg "/> < video aria-label ="Close Laptop: Inpainting · ProPainter " class ="poster-pending " controls ="" data-src ="assets/videos/visual-laptop-3.mp4 " loop ="" muted ="" playsinline ="" poster ="assets/images/visual-laptop-3.jpg " preload ="none "> </ video > </ div > < figcaption > < span class ="workflow-number "> 03</ span > Inpainting · ProPainter</ figcaption > </ figure > < figure class ="media-card matrix-card "> < div class ="video-frame poster-shell "> < img alt ="" aria-hidden ="true " class ="video-thumbnail " decoding ="async " loading ="lazy " src ="assets/images/visual-laptop-4.jpg "/> < video aria-label ="Close Laptop: Robot-arm rendering " class ="poster-pending " controls ="" data-src ="assets/videos/visual-laptop-4.mp4 " loop ="" muted ="" playsinline ="" poster ="assets/images/visual-laptop-4.jpg " preload ="none "> </ video > </ div > < figcaption > < span class ="workflow-number "> 04</ span > Robot-arm rendering</ figcaption > </ figure > </ div > </ div > < div class ="task-matrix-row visual-workflow " data-matrix-row ="visual-fridge " data-sync ="true "> < div class ="matrix-heading "> < h3 > Open Fridge</ h3 > < button aria-pressed ="false " class ="button small matrix-play " type ="button "> Play workflow</ button > </ div > < div class ="task-matrix-grid "> < figure class ="media-card matrix-card "> < div class ="video-frame poster-shell "> < img alt ="" aria-hidden ="true " class ="video-thumbnail " decoding ="async " loading ="lazy " src ="assets/images/visual-fridge-1.jpg "/> < video aria-label ="Open Fridge: Human RGB " class ="poster-pending " controls ="" data-src ="assets/videos/visual-fridge-1.mp4 " loop ="" muted ="" playsinline ="" poster ="assets/images/visual-fridge-1.jpg " preload ="none "> </ video > </ div > < figcaption > < span class ="workflow-number "> 01</ span > Human RGB</ figcaption > </ figure > < figure class ="media-card matrix-card "> < div class ="video-frame poster-shell "> < img alt ="" aria-hidden ="true " class ="video-thumbnail " decoding ="async " loading ="lazy " src ="assets/images/visual-fridge-2.jpg "/> < video aria-label ="Open Fridge: Arm-removal mask " class ="poster-pending " controls ="" data-src ="assets/videos/visual-fridge-2.mp4 " loop ="" muted ="" playsinline ="" poster ="assets/images/visual-fridge-2.jpg " preload ="none "> </ video > </ div > < figcaption > < span class ="workflow-number "> 02</ span > Arm-removal mask</ figcaption > </ figure > < figure class ="media-card matrix-card "> < div class ="video-frame poster-shell "> < img alt ="" aria-hidden ="true " class ="video-thumbnail " decoding ="async " loading ="lazy " src ="assets/images/visual-fridge-3.jpg "/> < video aria-label ="Open Fridge: Inpainting · DiffuEraser " class ="poster-pending " controls ="" data-src ="assets/videos/visual-fridge-3.mp4 " loop ="" muted ="" playsinline ="" poster ="assets/images/visual-fridge-3.jpg " preload ="none "> </ video > </ div > < figcaption > < span class ="workflow-number "> 03</ span > Inpainting · DiffuEraser</ figcaption > </ figure > < figure class ="media-card matrix-card "> < div class ="video-frame poster-shell "> < img alt ="" aria-hidden ="true " class ="video-thumbnail " decoding ="async " loading ="lazy " src ="assets/images/visual-fridge-4.jpg "/> < video aria-label ="Open Fridge: Robot-arm rendering " class ="poster-pending " controls ="" data-src ="assets/videos/visual-fridge-4.mp4 " loop ="" muted ="" playsinline ="" poster ="assets/images/visual-fridge-4.jpg " preload ="none "> </ video > </ div > < figcaption > < span class ="workflow-number "> 04</ span > Robot-arm rendering</ figcaption > </ figure > </ div > </ div > < div class ="task-matrix-row visual-workflow " data-matrix-row ="visual-curtain " data-sync ="true "> < div class ="matrix-heading "> < h3 > Pull Curtain</ h3 > < button aria-pressed ="false " class ="button small matrix-play " type ="button "> Play workflow</ button > </ div > < div class ="task-matrix-grid "> < figure class ="media-card matrix-card "> < div class ="video-frame poster-shell "> < img alt ="" aria-hidden ="true " class ="video-thumbnail " decoding ="async " loading ="lazy " src ="assets/images/visual-curtain-1.jpg "/> < video aria-label ="Pull Curtain: Human RGB " class ="poster-pending " controls ="" data-src ="assets/videos/visual-curtain-1.mp4 " loop ="" muted ="" playsinline ="" poster ="assets/images/visual-curtain-1.jpg " preload ="none "> </ video > </ div > < figcaption > < span class ="workflow-number "> 01</ span > Human RGB</ figcaption > </ figure > < figure class ="media-card matrix-card "> < div class ="video-frame poster-shell "> < img alt ="" aria-hidden ="true " class ="video-thumbnail " decoding ="async " loading ="lazy " src ="assets/images/visual-curtain-2.jpg "/> < video aria-label ="Pull Curtain: Arm-removal mask " class ="poster-pending " controls ="" data-src ="assets/videos/visual-curtain-2.mp4 " loop ="" muted ="" playsinline ="" poster ="assets/images/visual-curtain-2.jpg " preload ="none "> </ video > </ div > < figcaption > < span class ="workflow-number "> 02</ span > Arm-removal mask</ figcaption > </ figure > < figure class ="media-card matrix-card "> < div class ="video-frame poster-shell "> < img alt ="" aria-hidden ="true " class ="video-thumbnail " decoding ="async " loading ="lazy " src ="assets/images/visual-curtain-3.jpg "/> < video aria-label ="Pull Curtain: Inpainting · DiffuEraser " class ="poster-pending " controls ="" data-src ="assets/videos/visual-curtain-3.mp4 " loop ="" muted ="" playsinline ="" poster ="assets/images/visual-curtain-3.jpg " preload ="none "> </ video > </ div > < figcaption > < span class ="workflow-number "> 03</ span > Inpainting · DiffuEraser</ figcaption > </ figure > < figure class ="media-card matrix-card "> < div class ="video-frame poster-shell "> < img alt ="" aria-hidden ="true " class ="video-thumbnail " decoding ="async " loading ="lazy " src ="assets/images/visual-curtain-4.jpg "/> < video aria-label ="Pull Curtain: Robot-arm rendering " class ="poster-pending " controls ="" data-src ="assets/videos/visual-curtain-4.mp4 " loop ="" muted ="" playsinline ="" poster ="assets/images/visual-curtain-4.jpg " preload ="none "> </ video > </ div > < figcaption > < span class ="workflow-number "> 04</ span > Robot-arm rendering</ figcaption > </ figure > </ div > </ div > < details class ="workflow-notes "> < summary > Workflow details</ summary > < p > Masks are refined from SAM2 guidance. Inpainting removes the human arms; glove restoration and G1 arm rendering produce the composite. ProPainter and DiffuEraser are alternative inpainting branches. These are qualitative workflow examples: the laptop sequence retains some background trails; the curtain composite is a candidate with visible wrist-tracker and proximal-arm limitations.</ p > </ details > < details class ="disclosure visual-evidence "> < summary > Visual alignment results < span aria-hidden ="true "> +</ span > </ summary > < div aria-label ="Visual alignment task scores " class ="table-scroll " role ="region " tabindex ="0 "> < table > < thead > < tr > < th scope ="col "> Training images</ th > < th scope ="col "> Laptop</ th > < th scope ="col "> Refrigerator</ th > < th scope ="col "> Curtain</ th > < th scope ="col "> Drawer</ th > </ tr > </ thead > < tbody >
93+ < div class ="task-matrix-row visual-workflow " data-matrix-row ="visual-laptop " data-sync ="true "> < div class ="matrix-heading "> < h3 > Close Laptop</ h3 > < button aria-pressed ="false " class ="button small matrix-play " type ="button "> Play workflow</ button > </ div > < div class ="task-matrix-grid "> < figure class ="media-card matrix-card "> < div class ="video-frame poster-shell "> < img alt ="" aria-hidden ="true " class ="video-thumbnail " decoding ="async " loading ="lazy " src ="assets/images/visual-laptop-1.jpg "/> < video aria-label ="Close Laptop: Human RGB " class ="poster-pending " controls ="" data-src ="assets/videos/visual-laptop-1.mp4 " loop ="" muted ="" playsinline ="" poster ="assets/images/visual-laptop-1.jpg " preload ="none "> </ video > </ div > < figcaption > < span class ="workflow-number "> 01</ span > Human RGB</ figcaption > </ figure > < figure class ="media-card matrix-card "> < div class ="video-frame poster-shell "> < img alt ="" aria-hidden ="true " class ="video-thumbnail " decoding ="async " loading ="lazy " src ="assets/images/visual-laptop-2.jpg "/> < video aria-label ="Close Laptop: Arm-removal mask " class ="poster-pending " controls ="" data-src ="assets/videos/visual-laptop-2.mp4 " loop ="" muted ="" playsinline ="" poster ="assets/images/visual-laptop-2.jpg " preload ="none "> </ video > </ div > < figcaption > < span class ="workflow-number "> 02</ span > Arm-removal mask</ figcaption > </ figure > < figure class ="media-card matrix-card "> < div class ="video-frame poster-shell "> < img alt ="" aria-hidden ="true " class ="video-thumbnail " decoding ="async " loading ="lazy " src ="assets/images/visual-laptop-3.jpg "/> < video aria-label ="Close Laptop: Inpainting · ProPainter " class ="poster-pending " controls ="" data-src ="assets/videos/visual-laptop-3.mp4 " loop ="" muted ="" playsinline ="" poster ="assets/images/visual-laptop-3.jpg " preload ="none "> </ video > </ div > < figcaption > < span class ="workflow-number "> 03</ span > Inpainting · ProPainter</ figcaption > </ figure > < figure class ="media-card matrix-card "> < div class ="video-frame poster-shell "> < img alt ="" aria-hidden ="true " class ="video-thumbnail " decoding ="async " loading ="lazy " src ="assets/images/visual-laptop-4-warm.jpg "/> < video aria-label ="Close Laptop: Robot-arm rendering " class ="poster-pending " controls ="" data-src ="assets/videos/visual-laptop-4-warm.mp4 " loop ="" muted ="" playsinline ="" poster ="assets/images/visual-laptop-4-warm.jpg " preload ="none "> </ video > </ div > < figcaption > < span class ="workflow-number "> 04</ span > Robot-arm rendering</ figcaption > </ figure > </ div > </ div > < div class ="task-matrix-row visual-workflow " data-matrix-row ="visual-fridge " data-sync ="true "> < div class ="matrix-heading "> < h3 > Open Fridge</ h3 > < button aria-pressed ="false " class ="button small matrix-play " type ="button "> Play workflow</ button > </ div > < div class ="task-matrix-grid "> < figure class ="media-card matrix-card "> < div class ="video-frame poster-shell "> < img alt ="" aria-hidden ="true " class ="video-thumbnail " decoding ="async " loading ="lazy " src ="assets/images/visual-fridge-1.jpg "/> < video aria-label ="Open Fridge: Human RGB " class ="poster-pending " controls ="" data-src ="assets/videos/visual-fridge-1.mp4 " loop ="" muted ="" playsinline ="" poster ="assets/images/visual-fridge-1.jpg " preload ="none "> </ video > </ div > < figcaption > < span class ="workflow-number "> 01</ span > Human RGB</ figcaption > </ figure > < figure class ="media-card matrix-card "> < div class ="video-frame poster-shell "> < img alt ="" aria-hidden ="true " class ="video-thumbnail " decoding ="async " loading ="lazy " src ="assets/images/visual-fridge-2.jpg "/> < video aria-label ="Open Fridge: Arm-removal mask " class ="poster-pending " controls ="" data-src ="assets/videos/visual-fridge-2.mp4 " loop ="" muted ="" playsinline ="" poster ="assets/images/visual-fridge-2.jpg " preload ="none "> </ video > </ div > < figcaption > < span class ="workflow-number "> 02</ span > Arm-removal mask</ figcaption > </ figure > < figure class ="media-card matrix-card "> < div class ="video-frame poster-shell "> < img alt ="" aria-hidden ="true " class ="video-thumbnail " decoding ="async " loading ="lazy " src ="assets/images/visual-fridge-3.jpg "/> < video aria-label ="Open Fridge: Inpainting · DiffuEraser " class ="poster-pending " controls ="" data-src ="assets/videos/visual-fridge-3.mp4 " loop ="" muted ="" playsinline ="" poster ="assets/images/visual-fridge-3.jpg " preload ="none "> </ video > </ div > < figcaption > < span class ="workflow-number "> 03</ span > Inpainting · DiffuEraser</ figcaption > </ figure > < figure class ="media-card matrix-card "> < div class ="video-frame poster-shell "> < img alt ="" aria-hidden ="true " class ="video-thumbnail " decoding ="async " loading ="lazy " src ="assets/images/visual-fridge-4.jpg "/> < video aria-label ="Open Fridge: Robot-arm rendering " class ="poster-pending " controls ="" data-src ="assets/videos/visual-fridge-4.mp4 " loop ="" muted ="" playsinline ="" poster ="assets/images/visual-fridge-4.jpg " preload ="none "> </ video > </ div > < figcaption > < span class ="workflow-number "> 04</ span > Robot-arm rendering</ figcaption > </ figure > </ div > </ div > < div class ="task-matrix-row visual-workflow " data-matrix-row ="visual-curtain " data-sync ="true "> < div class ="matrix-heading "> < h3 > Pull Curtain</ h3 > < button aria-pressed ="false " class ="button small matrix-play " type ="button "> Play workflow</ button > </ div > < div class ="task-matrix-grid "> < figure class ="media-card matrix-card "> < div class ="video-frame poster-shell "> < img alt ="" aria-hidden ="true " class ="video-thumbnail " decoding ="async " loading ="lazy " src ="assets/images/visual-curtain-1.jpg "/> < video aria-label ="Pull Curtain: Human RGB " class ="poster-pending " controls ="" data-src ="assets/videos/visual-curtain-1.mp4 " loop ="" muted ="" playsinline ="" poster ="assets/images/visual-curtain-1.jpg " preload ="none "> </ video > </ div > < figcaption > < span class ="workflow-number "> 01</ span > Human RGB</ figcaption > </ figure > < figure class ="media-card matrix-card "> < div class ="video-frame poster-shell "> < img alt ="" aria-hidden ="true " class ="video-thumbnail " decoding ="async " loading ="lazy " src ="assets/images/visual-curtain-2.jpg "/> < video aria-label ="Pull Curtain: Arm-removal mask " class ="poster-pending " controls ="" data-src ="assets/videos/visual-curtain-2.mp4 " loop ="" muted ="" playsinline ="" poster ="assets/images/visual-curtain-2.jpg " preload ="none "> </ video > </ div > < figcaption > < span class ="workflow-number "> 02</ span > Arm-removal mask</ figcaption > </ figure > < figure class ="media-card matrix-card "> < div class ="video-frame poster-shell "> < img alt ="" aria-hidden ="true " class ="video-thumbnail " decoding ="async " loading ="lazy " src ="assets/images/visual-curtain-3.jpg "/> < video aria-label ="Pull Curtain: Inpainting · DiffuEraser " class ="poster-pending " controls ="" data-src ="assets/videos/visual-curtain-3.mp4 " loop ="" muted ="" playsinline ="" poster ="assets/images/visual-curtain-3.jpg " preload ="none "> </ video > </ div > < figcaption > < span class ="workflow-number "> 03</ span > Inpainting · DiffuEraser</ figcaption > </ figure > < figure class ="media-card matrix-card "> < div class ="video-frame poster-shell "> < img alt ="" aria-hidden ="true " class ="video-thumbnail " decoding ="async " loading ="lazy " src ="assets/images/visual-curtain-4.jpg "/> < video aria-label ="Pull Curtain: Robot-arm rendering " class ="poster-pending " controls ="" data-src ="assets/videos/visual-curtain-4.mp4 " loop ="" muted ="" playsinline ="" poster ="assets/images/visual-curtain-4.jpg " preload ="none "> </ video > </ div > < figcaption > < span class ="workflow-number "> 04</ span > Robot-arm rendering</ figcaption > </ figure > </ div > </ div > < details class ="workflow-notes "> < summary > Workflow details</ summary > < p > Masks are refined from SAM2 guidance. Inpainting removes the human arms; glove restoration and G1 arm rendering produce the composite. ProPainter and DiffuEraser are alternative inpainting branches. These are qualitative workflow examples: the laptop sequence retains some background trails; the curtain composite is a candidate with visible wrist-tracker and proximal-arm limitations.</ p > </ details > < details class ="disclosure visual-evidence "> < summary > Visual alignment results < span aria-hidden ="true "> +</ span > </ summary > < div aria-label ="Visual alignment task scores " class ="table-scroll " role ="region " tabindex ="0 "> < table > < thead > < tr > < th scope ="col "> Training images</ th > < th scope ="col "> Laptop</ th > < th scope ="col "> Refrigerator</ th > < th scope ="col "> Curtain</ th > < th scope ="col "> Drawer</ th > </ tr > </ thead > < tbody >
9494< tr > < th scope ="row "> Human RGB</ th > < td > 31.67</ td > < td > 16.67</ td > < td > 35.00</ td > < td > 50.00</ td > </ tr >
9595< tr > < th scope ="row "> + Arm rendering</ th > < td > 51.67</ td > < td > 23.33</ td > < td > 41.67</ td > < td > —</ td > </ tr >
9696< tr > < th scope ="row "> + Augmentation</ th > < td > 61.67</ td > < td > 33.33</ td > < td > 33.33</ td > < td > 61.67</ td > </ tr >
You can’t perform that action at this time.
0 commit comments