Hello, I have some questions about neural network processing with YOLO... I have a scenario where I need to identify coffee boxes whose only difference is the color to identify one type of coffee as different from another. I have always believed that for performance reasons, network training converts images to grayscale to work on only 1 color channel and does the same when inferring new images. Is this understanding correct? Is it possible to train the models considering the colors, the 3 RGB channels? Or would the best option be to identify the objects (bounding boxes) and perform post-processing with OpenCV, for example, in the bouding box region to identify the closest color? Thanks

Would this be good for hate meme or offensive meme pictures detection? Memes that have some pictures along with text in them????

Toy Plains

insanely goated bro

TypeError: unhashable type: 'numpy.ndarray' (Having this error in custom training process)

A good question would be whether open AI is ever going to contribute anything real to Ai research, theory, and progress. So far, all they've done is take somebody else's idea - using attention with a pre-trained transformer - and milk it endlessly without any transformative ideas of their own. They have not actually changed the state of the art at all, just re-engineered it and threw ever more resources at it. There is no fundamental difference between GPT 4o and GPT 2. They just keep throwing memory, data, and RAG engineering plug-in gimmicks at it.

Taylor Michelle Lopez Lisa Lewis Cynthia

Is Annotation Necessary for the Model to Learn the Details in the Image? or beneficial ? for example, when I throw an image and ask, there is a speed sign on the left side, there are children eating ice cream on the right side, the stork is flying above, the weather is sunny, the ground is asphalt and pavement, etc. Do I need to anotate these for the model to learn them?

Awesome video. Could you please provide the code to download the final video outputs? Also, how do you overlay the voronoi diagram and the radar style view onto the final output ? I tried to rewrite the code below based on your code from the beginning but the ellipses, labels and triangle don't track the players and the ball correctly. I'm almost there, please advise, thanks! video_info = sv.VideoInfo.from_video_path(SOURCE_VIDEO_PATH) video_sink = sv.VideoSink(TARGET_VIDEO_PATH, video_info=video_info) frame_generator = sv.get_video_frames_generator(SOURCE_VIDEO_PATH) with video_sink: for frame in tqdm(frame_generator, total=video_info.total_frames): result = PLAYER_DETECTION_MODEL.infer(frame, confidence=0.3)[0] detections = sv.Detections.from_inference(result) labels = [ f"#{tracker_id}" for tracker_id in all_detections.tracker_id ] all_detections.class_id = all_detections.class_id.astype(int) annotated_frame = frame.copy() annotated_frame = ellipse_annotator.annotate( scene=annotated_frame, detections=all_detections) annotated_frame = label_annotator.annotate( scene=annotated_frame, detections=all_detections, labels=labels) annotated_frame = triangle_annotator.annotate( scene=annotated_frame, detections=ball_detections, ) video_sink.write_frame(annotated_frame)

Awesome video. Could you please provide the code to download the final video outputs? Also, how do you overlay the voronoi diagram and the radar style view onto the final output ? I tried to rewrite the code below based on your code from the beginning but the ellipses, labels and triangle don't track the players and the ball correctly. I'm almost there, please advise, thanks! video_info = sv.VideoInfo.from_video_path(SOURCE_VIDEO_PATH) video_sink = sv.VideoSink(TARGET_VIDEO_PATH, video_info=video_info) frame_generator = sv.get_video_frames_generator(SOURCE_VIDEO_PATH) with video_sink: for frame in tqdm(frame_generator, total=video_info.total_frames): result = PLAYER_DETECTION_MODEL.infer(frame, confidence=0.3)[0] detections = sv.Detections.from_inference(result) labels = [ f"#{tracker_id}" for tracker_id in all_detections.tracker_id ] all_detections.class_id = all_detections.class_id.astype(int) annotated_frame = frame.copy() annotated_frame = ellipse_annotator.annotate( scene=annotated_frame, detections=all_detections) annotated_frame = label_annotator.annotate( scene=annotated_frame, detections=all_detections, labels=labels) annotated_frame = triangle_annotator.annotate( scene=annotated_frame, detections=ball_detections, ) video_sink.write_frame(annotated_frame)

Can you download the trained model (weigjts) ? i want to get the .pt file (for pytorch)

Jackson Christopher White Elizabeth Walker Brian

Tutorial starts at 14:21

Next, plan for llama vlm fine tuning with roboflow datasets, probably object detection

Hey thanks for the Tutorial! I tried to follow your steps, but i've got an error at uploading my trained model to roboflow at 11:30. The message is: An error occured when getting the model upload URL: 404 Client Error: Not Found for url: ... Any ideas?

how he said - easy raid?

So what if I dont have a model, what do I do? could you please help me

FUTBol

What is the object detection inference latency with this comparable to YOLO-world?

bro can I use this to make a football game. I am not a developer of any kind, I want a full realistic sim football game

Hi can anyone tell me how to chose which yaml file is best suitable for our model?

I want to put a camera at the front door and identify visitors - my family/stranger. What is the best way to do this?

What's the secret roboflow key

Bethel Junctions

Your video provides valuable insights into computer vision, and we truly appreciate the depth of information you share. it helps many of us learn and grow. Thank you for your dedication and efforts! 😊

Crushing it Piotr!!!!!

I am working on a project and considering using Roboflow. In my project, I will classify the geometric shapes (triangle, quadrilateral, circle, zigzag, etc.) found on shoe soles. At this point, I will label them using object detection. What is your suggestion on this?

ipad is not supported in using cvat 😢

How can I then deploy the trained model on roboflow pls ? It is said it is not supported and if i canhed to yolov8 there is a depedency issue : ultralytics==8.0.196 is required whereas ultralytics==8.3.2 is installed

Was something in Roboflow package updated recently? Getting an error trying to download the keypoint dataset. RuntimeError Traceback (most recent call last) <ipython-input-5-0647745e9d8d> in <cell line: 9>() 7 project = rf.workspace("roboflow-jvuqo").project("football-field-detection-f07vi") 8 version = project.version(12) ----> 9 dataset = version.download("yolov8") /usr/local/lib/python3.10/dist-packages/roboflow/core/version.py in download(self, model_format, location, overwrite) 241 else: 242 try: --> 243 raise RuntimeError(response.json()) 244 except json.JSONDecodeError: 245 response.raise_for_status() RuntimeError: {'error': 'Unknown error'}

Hi, Thanks for your very nice video. I see you are using RTMDET from mmyolo (not from mmdetection). So, I was wondering if it is still free to use it for commercial. I mean the licence of mmyolo is different than mmdetection, and you used mmyolo model in your video. So, has RTMDET from mmyolo the same licence of RTMDET from mmdetection which was "Apache License 2.0"?

How would i go about measuring the duration of each class that appears in a zone.

At which point do I have to start watching to skip over the training part? I am getting confused because the material is very difficult to understand. Also have you provided all the files and resources to start off from that point?

Can I use Yolo11 for tracking

As someone who is at the starting gates for YOLOv8 and Roboflow, many, many thanks for publishing these tutorials. You not only keep the presentation at a level that is easily understood (by a dinosaur like me from the punched cards era) but also make it exciting to try out the applications at our own pace owing to superb post-publication support. Best regards.

Breaking News: YOLO v11 Outperforms Previous Versions in Object Detection! Please see Comparision of YOLO v11, YOLO v10, and YOLO v8 across various metrics, including mAP, parameters, and real-world performance. Key Findings: Superior Accuracy: YOLO v11 demonstrated significantly higher accuracy compared to its predecessors on both COCO and OBB-Dota V1 datasets. Check YOLO v11 video tutorial and Colab Demo, where they showcase live demos and discuss the results obtained with YOLO Nano on satellite data. kzbin.info/www/bejne/g52ZnKefotN9gqM #YOLO #deeplearning #objectdetection #yolo_v11 #yolo_v10 #yolo_v8 #computervision #innovation

i am getting error again and again if i use any other dataset , my epochs are not running only.. suggest what to do

Thank you so much for stream 🎉😊

Hi, can I input a video file of a full match and let it generate a video containing all touches by a certain player?

I get below error when trying time-in-zone... ARN:[email protected]] global cap_msmf.cpp:488 `anonymous-namespace'::SourceReaderCB::OnReadSample videoio(MSMF): async ReadSample() call is failed with error status: -1072875772 [ WARN:[email protected]] global cap_msmf.cpp:1769 CvCapture_MSMF::grabFrame videoio(MSMF): can't grab frame. Error: -1072875772

i lost it when "it's called football"😂THAT'S RIGHT BROOOO!

Martinez Larry Martin Edward Clark Kimberly

Hello, are you able to make an updated guide regarding training custom datasets into the EfficientDet model? It seems like most functions have been deprecated. Thank you in advance.

Meghan Stravenue

I encounter an error while trying to continue deploying the trained model with the current version of ultralytics (8.2.103) instead of the said dependency (8.0.196). Suggestions?

Hi Is it possible to get the GitHub repo or other code associated with the working of Florence

our project is to detect and count the object on the captured photo. can we follow this tutorial? or is there other more applicable tutorial we can follow

how much will it cost for a single stream for a month with maybe 2 fps provided we have already trained the model with roboflow?

Hey man, i'm back here! Can you help me turning the Radar-view and Voronoi Diagram into video ? There's a long time that i'm trying it, but it always generates a 258Bytes video...