https://spikedge.com/tr/videolar/her-seyi-hesaplamak/ https://spikedge.com/media/films/compute-everything.jpg Her şeyi hesaplamak mı? Kestirimci bakımda nöromorfik işlemci ve GPU Uçta asıl soru ne kadar hızlı hesapladığımız değil, her şeyi hesaplamamız gerekip gerekmediği. Jetson Orin Nano Super her sensör penceresini modelin tamamından geçirir; BrainChip AKD1500 yalnızca bir olay geldiğinde çalışır. Örnek bir pompa motorunun rulmanı. https://www.youtube.com/embed/00fZXpj2zuQ 60 2026-10-18T02:22:46-07:00 yes no no https://spikedge.com/tr/videolar/dogrudan-gpuya/ https://spikedge.com/media/films/straight-to-gpu.jpg Doğrudan GPU'ya: TI Jacinto'dan Jetson Orin'e zero-copy PCIe Tek kamera karesi, dört GPU ve onu görmeyi bekleyen bir operatör. Kare TI Jacinto'ya iner, oradan sonrasını DMA üstlenir ve PCIe üzerinden doğrudan Jetson Orin'in GPU belleğine yazılır; işlemci kareye dokunmaz. Sonra aynı kartlar ve aynı kablo, bu kez her kareyi işlemci kopyalarken. https://www.youtube.com/embed/pbGVIkvVZak 64 2026-09-28T08:16:14-07:00 yes no no https://spikedge.com/tr/videolar/raylar-cetveldir/ https://spikedge.com/media/films/rails-ruler.jpg Demiryolu engel tespiti için sentetik veri: raylar cetveldir Raydaki engeli ne bekleyebilirsiniz ne sahneleyebilirsiniz. Bu yüzden eğitim verisi gerçek kabin görüntüsü üzerine kuruluyor ve her adım ölçülüyor: raylar kalibrasyonsuz ufku ve ölçeği veriyor, yerleştirilen engel sahnede sabit kalıyor, etiketi de görünen kısmı izliyor. https://www.youtube.com/embed/4Zf0l2JFHJU 60 2026-09-26T08:21:39-07:00 yes no no https://spikedge.com/tr/videolar/gercek-zamanin-bir-bedeli-var/ https://spikedge.com/media/films/real-time-cost.jpg Gerçek zamanın bir bedeli var: ne zaman hesaplamamalı Direkteki bir sensör gökyüzündeki bir nesneyi izliyor. Ağır model tek bir karede koşuyor, aradaki kareleri çok daha ucuz bir tahmin taşıyor. Kesintisiz farkındalık, gereksiz hesap olmadan. https://www.youtube.com/embed/jcUtHuWG7OM 61 2026-09-22T03:23:17-07:00 yes no no https://spikedge.com/tr/videolar/aracinizin-goremedigi-nesne/ https://spikedge.com/media/films/cooperative-perception.jpg Aracınızın göremediği nesne: işbirlikçi algı Otonom bir terminal traktörü, konteyner istifinin bir forklifti gizlediği kavşağa yaklaşıyor. Yol kenarındaki algı düğümü bir video akışı daha değil nesnenin durumunu gönderiyor; araç kendi gözüyle görmeden yavaşlıyor. https://www.youtube.com/embed/Dj9kEgYOmnk 96 2026-09-06T02:45:37-07:00 yes no no https://spikedge.com/tr/videolar/iz-zamandaki-bir-sekildir/ https://spikedge.com/media/films/sky-track.jpg İz, zamanda bir şekildir: tek direkte kuşu drone'dan ayırmak Tek karede kuş ile drone aynı koyu noktadır. İzleyici her noktayı bir yola çevirir; görüntüdeki izlerden biri hiç çırpmaz, düz gider, havada durur ve dönmeden yana kayar. İkinci ipucu ses. https://www.youtube.com/embed/-MSIGtGuxII 80 2026-08-31T02:59:32-07:00 yes no no https://spikedge.com/tr/videolar/tek-cerceve-butun-makineler-ethercat/ https://spikedge.com/media/films/ethercat-plant.jpg Tek çerçeve, bütün makineler: EtherCAT veriyi üretim hattında nasıl taşır Kontrolcü her çevrimde, hattaki her cihaz için yeri olan tek bir çerçeve kurar. Çerçeve geçerken her cihaz kendi komutunu alır ve cevabını geri yazar; donanımda, yolda. Tezgâhımızda ölçülen: ortalama 19,2 µs, en fazla 30,9 µs, %0 paket kaybı. https://www.youtube.com/embed/IzI3t_eqH1g 80 2026-08-26T02:15:26-07:00 yes no no https://spikedge.com/tr/videolar/kisi-beklemedik-tek-sahne-dort-kosul/ https://spikedge.com/media/films/rail-conditions.jpg Kışı beklemedik: tek demiryolu sahnesi, dört koşul (sentetik veri) Tek demiryolu sahnesi, tek kamera yolu, dört çevre koşulu; çekilmedi, Blender'da render edildi. Turuncu kutular renderer'ın yazdığı yer gerçeğidir ve dört koşulda da aynıdır. https://www.youtube.com/embed/4L9Oa1-V09c 27 2026-08-16T02:40:00-07:00 yes no no https://spikedge.com/tr/videolar/gercek-ray-goruntusunde-otomatik-etiketli-engeller/ https://spikedge.com/media/films/rail-autolabel.jpg Gerçek demiryolu görüntüsünde otomatik etiketli sentetik engeller Gerçek kabin görüntüsü üzerine sentetik engel: her karede ray segmentasyonu, ray açıklığına göre ölçeklenen ve dünyaya kilitlenen yerleşim, yerleştirme adımından çıkan YOLO etiketleri. Bu koşu: 960 kare, 8 sınıf, 3.145 etiket. https://www.youtube.com/embed/oph9tXZgA-Y 20 2026-08-08T23:14:27-07:00 yes no no https://spikedge.com/tr/videolar/cekemeyeceginiz-egitim-verisi/ https://spikedge.com/media/films/rail-synthetic.jpg Çekemeyeceğiniz eğitim verisi: Blender'da sentetik demiryolu engelleri Eğitim verisi toplamak için bir trenin önüne kaya koyamazsınız. Sahne Blender'da üretiliyor ve etiketler onunla birlikte çıkıyor: raydaki kaya, hayvan ve kutu, her biri otomatik kutusu ve sınıf etiketiyle. https://www.youtube.com/embed/op-CQgPbATk 35 2026-08-05T12:03:35-07:00 yes no no https://spikedge.com/en/videos/compute-everything/ https://spikedge.com/media/films/compute-everything.jpg Compute Everything? Neuromorphic vs GPU for predictive maintenance At the edge, the real question isn't how fast we can compute. It's whether we need to compute it all. A Jetson Orin Nano Super runs every sensor window through the whole model; a BrainChip AKD1500 works only when an event arrives. The example is a bearing on a pump motor. https://www.youtube.com/embed/00fZXpj2zuQ 60 2026-10-18T02:22:46-07:00 yes no no https://spikedge.com/en/videos/straight-to-the-gpu/ https://spikedge.com/media/films/straight-to-gpu.jpg Straight to the GPU: zero-copy PCIe from TI Jacinto to Jetson Orin One camera frame, four GPUs, and an operator waiting to see it. The frame lands on a TI Jacinto, DMA takes it from there, and it crosses PCIe straight into the GPU memory of a Jetson Orin; the CPU never touches it. Then the same boards and the same cable, with the CPU copying every frame. https://www.youtube.com/embed/pbGVIkvVZak 64 2026-09-28T08:16:14-07:00 yes no no https://spikedge.com/en/videos/the-rails-are-the-ruler/ https://spikedge.com/media/films/rails-ruler.jpg Synthetic data for railway obstacle detection: the rails are the ruler You cannot wait for an obstacle on the track, and you cannot stage one. So the training data is built on real cab-view video, and every step is measured: the rails give horizon and scale without calibration, a placed obstacle stays fixed in the scene, and its label follows what stays visible. https://www.youtube.com/embed/4Zf0l2JFHJU 60 2026-09-26T08:21:39-07:00 yes no no https://spikedge.com/en/videos/real-time-has-a-cost/ https://spikedge.com/media/films/real-time-cost.jpg Real time has a cost: when not to compute A sensor on a mast keeps track of an object in the sky. The heavy model runs on one frame, and the frames in between are carried forward by a far cheaper prediction. Continuous awareness, less unnecessary compute. https://www.youtube.com/embed/jcUtHuWG7OM 61 2026-09-22T03:23:17-07:00 yes no no https://spikedge.com/en/videos/the-object-your-vehicle-cannot-see/ https://spikedge.com/media/films/cooperative-perception.jpg The object your vehicle cannot see: cooperative perception An autonomous terminal tractor approaches a junction where a container stack hides a forklift. A roadside perception node sends the state of the object, not another video stream, and the vehicle slows before it can see anything itself. https://www.youtube.com/embed/Dj9kEgYOmnk 96 2026-09-06T02:45:37-07:00 yes no no https://spikedge.com/en/videos/a-track-is-a-shape-in-time/ https://spikedge.com/media/films/sky-track.jpg A track is a shape in time: telling a bird from a drone on one mast In a single frame a bird and a drone are the same dark dot. A tracker turns each dot into a path; one path in this film never flaps, flies straight, stops in the air and slides sideways without turning. Sound is the second cue. https://www.youtube.com/embed/-MSIGtGuxII 80 2026-08-31T02:59:32-07:00 yes no no https://spikedge.com/en/videos/one-frame-every-machine-ethercat/ https://spikedge.com/media/films/ethercat-plant.jpg One frame, every machine: how EtherCAT moves data on a production line The controller builds one frame every cycle, with a place for every device on the line. As it passes, each device takes its command out and writes its answer back in, in hardware, in transit. Measured on our bench: 19.2 µs average, 30.9 µs maximum, 0% packet loss. https://www.youtube.com/embed/IzI3t_eqH1g 80 2026-08-26T02:15:26-07:00 yes no no https://spikedge.com/en/videos/we-didnt-wait-for-winter/ https://spikedge.com/media/films/rail-conditions.jpg We didn't wait for winter: one railway scene, four conditions (synthetic data) One railway scene, one camera path, four environment conditions, rendered in Blender instead of filmed. The orange boxes are ground truth written by the renderer, identical in all four conditions. https://www.youtube.com/embed/4L9Oa1-V09c 27 2026-08-16T02:40:00-07:00 yes no no https://spikedge.com/en/videos/auto-labeled-synthetic-obstacles-on-real-railway-footage/ https://spikedge.com/media/films/rail-autolabel.jpg Auto-labeled synthetic obstacles on real railway footage Synthetic obstacles composited onto real cab-view footage: rail segmentation per frame, world-locked placement scaled against the track gauge, and YOLO labels that come out of the placement step. This run: 960 frames, 8 classes, 3,145 annotations. https://www.youtube.com/embed/oph9tXZgA-Y 20 2026-08-08T23:14:27-07:00 yes no no https://spikedge.com/en/videos/training-data-you-cannot-film/ https://spikedge.com/media/films/rail-synthetic.jpg Training data you cannot film: synthetic railway obstacles in Blender You cannot put a rock in front of a train to collect training data. The scene is generated in Blender instead, and the labels come out with it: rock, animal and box on the track, each with an automatically generated box and class label. https://www.youtube.com/embed/op-CQgPbATk 35 2026-08-05T12:03:35-07:00 yes no no https://spikedge.com/de/videos/alles-berechnen/ https://spikedge.com/media/films/compute-everything.jpg Alles berechnen? Neuromorpher Prozessor und GPU in der vorausschauenden Wartung An der Edge ist die eigentliche Frage nicht, wie schnell wir rechnen, sondern ob wir alles rechnen müssen. Ein Jetson Orin Nano Super schickt jedes Sensorfenster durch das ganze Modell; ein BrainChip AKD1500 arbeitet nur, wenn ein Ereignis eintrifft. Das Beispiel ist ein Lager an einem Pumpenmotor. https://www.youtube.com/embed/00fZXpj2zuQ 60 2026-10-18T02:22:46-07:00 yes no no https://spikedge.com/de/videos/direkt-in-die-gpu/ https://spikedge.com/media/films/straight-to-gpu.jpg Direkt in die GPU: Zero-Copy-PCIe von TI Jacinto zu Jetson Orin Ein Kamerabild, vier GPUs und ein Bediener, der darauf wartet. Das Bild landet auf einem TI Jacinto, von dort übernimmt DMA, und es geht per PCIe direkt in den GPU-Speicher eines Jetson Orin; die CPU fasst es nicht an. Danach dieselben Boards und dasselbe Kabel, diesmal kopiert die CPU jedes Bild. https://www.youtube.com/embed/pbGVIkvVZak 64 2026-09-28T08:16:14-07:00 yes no no https://spikedge.com/de/videos/die-schienen-sind-das-lineal/ https://spikedge.com/media/films/rails-ruler.jpg Synthetische Daten für die Hinderniserkennung am Gleis: Die Schienen sind das Lineal Auf ein Hindernis im Gleis kann man nicht warten, und man kann keines nachstellen. Deshalb entstehen die Trainingsdaten auf realen Führerstandsaufnahmen, und jeder Schritt wird gemessen: Die Schienen liefern Horizont und Maßstab ohne Kalibrierung, ein platziertes Hindernis bleibt in der Szene fest, sein Label folgt dem sichtbaren Teil. https://www.youtube.com/embed/4Zf0l2JFHJU 60 2026-09-26T08:21:39-07:00 yes no no https://spikedge.com/de/videos/echtzeit-hat-ihren-preis/ https://spikedge.com/media/films/real-time-cost.jpg Echtzeit hat ihren Preis: wann man nicht rechnen sollte Ein Sensor am Mast verfolgt ein Objekt am Himmel. Das schwere Modell läuft auf einem Frame, die Frames dazwischen trägt eine weit günstigere Vorhersage weiter. Durchgehende Wahrnehmung, weniger unnötige Rechenlast. https://www.youtube.com/embed/jcUtHuWG7OM 61 2026-09-22T03:23:17-07:00 yes no no https://spikedge.com/de/videos/das-objekt-das-ihr-fahrzeug-nicht-sieht/ https://spikedge.com/media/films/cooperative-perception.jpg Das Objekt, das Ihr Fahrzeug nicht sieht: kooperative Wahrnehmung Ein autonomer Terminal-Trecker nähert sich einer Kreuzung, an der ein Containerstapel einen Stapler verdeckt. Ein Wahrnehmungsknoten am Fahrbahnrand sendet den Zustand des Objekts statt eines weiteren Videostreams, und das Fahrzeug verzögert, bevor es selbst etwas sieht. https://www.youtube.com/embed/Dj9kEgYOmnk 96 2026-09-06T02:45:37-07:00 yes no no https://spikedge.com/de/videos/eine-spur-ist-eine-form-in-der-zeit/ https://spikedge.com/media/films/sky-track.jpg Eine Spur ist eine Form in der Zeit: Vogel oder Drohne an einem Mast In einem einzelnen Frame sind Vogel und Drohne derselbe dunkle Punkt. Ein Tracker macht aus jedem Punkt einen Weg; eine Spur in diesem Film schlägt nie mit den Flügeln, fliegt geradeaus, steht in der Luft und gleitet seitwärts, ohne sich zu drehen. Der Ton ist der zweite Hinweis. https://www.youtube.com/embed/-MSIGtGuxII 80 2026-08-31T02:59:32-07:00 yes no no https://spikedge.com/de/videos/ein-frame-jede-maschine-ethercat/ https://spikedge.com/media/films/ethercat-plant.jpg Ein Frame, jede Maschine: wie EtherCAT Daten in der Fertigungslinie bewegt Der Controller baut in jedem Zyklus einen Frame mit einem Platz für jedes Gerät der Linie. Im Durchlauf entnimmt jedes Gerät seinen Befehl und schreibt seine Antwort zurück, in Hardware, unterwegs. Auf unserem Prüfstand gemessen: 19,2 µs im Mittel, 30,9 µs maximal, 0 % Paketverlust. https://www.youtube.com/embed/IzI3t_eqH1g 80 2026-08-26T02:15:26-07:00 yes no no https://spikedge.com/de/videos/wir-haben-nicht-auf-den-winter-gewartet/ https://spikedge.com/media/films/rail-conditions.jpg Wir haben nicht auf den Winter gewartet: eine Bahnszene, vier Bedingungen (synthetische Daten) Eine Bahnszene, ein Kamerapfad, vier Umgebungsbedingungen, in Blender gerendert statt gefilmt. Die orangefarbenen Boxen sind vom Renderer geschriebene Ground Truth, in allen vier Bedingungen identisch. https://www.youtube.com/embed/4L9Oa1-V09c 27 2026-08-16T02:40:00-07:00 yes no no https://spikedge.com/de/videos/automatisch-annotierte-hindernisse-auf-realen-bahnaufnahmen/ https://spikedge.com/media/films/rail-autolabel.jpg Automatisch annotierte synthetische Hindernisse auf realen Bahnaufnahmen Synthetische Hindernisse auf realen Führerstandsaufnahmen: Schienensegmentierung pro Frame, weltfeste, an der Spurweite skalierte Platzierung und YOLO-Labels, die aus dem Platzierungsschritt entstehen. Dieser Lauf: 960 Frames, 8 Klassen, 3.145 Annotationen. https://www.youtube.com/embed/oph9tXZgA-Y 20 2026-08-08T23:14:27-07:00 yes no no https://spikedge.com/de/videos/trainingsdaten-die-man-nicht-filmen-kann/ https://spikedge.com/media/films/rail-synthetic.jpg Trainingsdaten, die man nicht filmen kann: synthetische Bahnhindernisse in Blender Man kann keinen Stein vor einen Zug legen, um Trainingsdaten zu sammeln. Die Szene wird stattdessen in Blender erzeugt, und die Labels entstehen mit: Stein, Tier und Kiste auf dem Gleis, jeweils mit automatisch erzeugter Box und Klasse. https://www.youtube.com/embed/op-CQgPbATk 35 2026-08-05T12:03:35-07:00 yes no no