{"id":"01aae0f1-f73a-4962-90ed-8497794836d1","arxiv_id":"2606.07659","paper_version":1,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":2.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":0,"one_line_summary":"Fine-tuned YOLOv8 with TensorRT/OpenVINO optimizations reaches over 120 FPS and 98.5% mAP on NEU, MVTec AD, and custom automotive defect datasets for edge deployment.","lead":"The paper fine-tunes YOLOv8 for spotting surface defects like scratches and pits in manufacturing and optimizes it with TensorRT and OpenVINO to run over 120 frames per second on edge hardware such as the NVIDIA Jetson Orin. A smart generalist might read it to learn how existing AI models can be adapted for real-time quality checks on factory floors without cloud delays.","discovery_kind":"new_application","skeptic_critique":{"model":"grok-4.3","headline":"Whether TensorRT/OpenVINO optimizations preserve the 98.5% mAP on the reported datasets without accuracy drop or post-hoc selection","rationale":"The reader's weakest assumption correctly isolates the single condition that must be true for the headline speed+accuracy claim to be credible. Because the original review was abstract-only, the full text could in principle supply the missing pre/post numbers and split details; until those are verified the verdict remains UNVERDICTED. No other internal inconsistency is visible from the supplied material.","tokens_in":1739,"tokens_out":376,"duration_ms":17070,"concrete_test":"From the results and experimental-setup sections, extract mAP@0.5:0.95 for (i) baseline YOLOv8, (ii) fine-tuned model before optimization, (iii) TensorRT/OpenVINO version on Jetson Orin, all on the same held-out test split; also record input resolution and batch size used for the 120 FPS measurement. If post-optimization mAP drops >1 point or the FPS figure uses a different resolution, the joint claim is unsupported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim requires that fine-tuning plus TensorRT/OpenVINO acceleration on Jetson Orin yields both >120 FPS and 98.5% mAP on NEU/MVTec/custom automotive data. This holds only if (a) mAP is measured on the final optimized model (not the unoptimized fine-tune), (b) no test-set leakage or selective reporting occurred during fine-tuning, and (c) the same input resolution and batch size are used for both accuracy and FPS numbers. The abstract supplies none of these controls; the full text must explicitly compare pre- and post-optimization mAP on identical splits.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The paper introduces Industrial-YOLO, a fine-tuned YOLOv8 architecture with TensorRT and OpenVINO optimizations for real-time surface defect detection on edge hardware. It reports systematic benchmarks on the NEU surface defect database, MVTec AD dataset, and custom automotive/battery manufacturing extensions, claiming inference exceeding 120 FPS on NVIDIA Jetson Orin while achieving 98.5% mAP.","tokens_in":1871,"tokens_out":340,"duration_ms":19475,"significance":"If the performance numbers are shown to hold under standard controls (identical splits for accuracy and speed, mAP measured on the final optimized model), the work would supply a practical reference implementation for AOI systems in manufacturing. The combination of public benchmarks with domain-specific extensions is a positive aspect.","major_comments":[{"comment":"Abstract: the headline claim of 98.5% mAP together with >120 FPS on the optimized Jetson Orin model cannot be evaluated because the text supplies no information on training/validation splits, hyperparameter selection, or whether mAP was computed on the TensorRT/OpenVINO model versus the unoptimized fine-tune.","section":"Abstract"},{"comment":"The central claim requires an explicit side-by-side comparison of mAP (and any other accuracy metrics) before and after TensorRT/OpenVINO optimization on the identical held-out test splits used for the reported 98.5% figure; without this comparison the preservation of accuracy under acceleration remains unverified.","section":"Experimental Results"}],"minor_comments":[],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for the constructive feedback highlighting the need for greater experimental transparency. We will revise the manuscript to address the points raised.","responses":[{"response":"We agree the manuscript should explicitly state these details. The revised version will update the abstract and add a methods subsection specifying the train/validation/test splits, hyperparameter selection process, and confirming that the 98.5% mAP was evaluated on the final TensorRT/OpenVINO-optimized model.","revision_made":"yes","referee_comment":"[Abstract] Abstract: the headline claim of 98.5% mAP together with >120 FPS on the optimized Jetson Orin model cannot be evaluated because the text supplies no information on training/validation splits, hyperparameter selection, or whether mAP was computed on the TensorRT/OpenVINO model versus the unoptimized fine-tune."},{"response":"We will add the requested side-by-side comparison in the experimental results section, reporting mAP and related metrics on the identical held-out test splits before and after optimization to verify accuracy preservation.","revision_made":"yes","referee_comment":"[Experimental Results] The central claim requires an explicit side-by-side comparison of mAP (and any other accuracy metrics) before and after TensorRT/OpenVINO optimization on the identical held-out test splits used for the reported 98.5% figure; without this comparison the preservation of accuracy under acceleration remains unverified."}],"tokens_in":1342,"tokens_out":324,"duration_ms":23433,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","letter":"The main takeaway is that the paper applies an existing model to industrial defect detection and reports strong speed and accuracy on edge hardware, but the supporting details are missing.\n\nWhat it does is benchmark YOLOv8 after fine-tuning on the NEU steel dataset and MVTec AD, then add custom examples from automotive and battery lines. The optimizations target the Jetson Orin and combine FPS with mAP, which at least addresses the practical constraints of real-time inspection.\n\nThe soft spots are more central. The abstract gives no training splits, hyperparameter choices, or validation protocol. It also does not confirm that the 98.5% mAP was measured after the TensorRT and OpenVINO steps rather than before, or that the same resolution and test set were used for both accuracy and speed. Without those checks, the central claim is difficult to evaluate. There are also no error bars or ablation results on how the custom data or the acceleration steps affect performance.\n\nThis paper is aimed at engineers who want a worked example of deploying object detection for manufacturing quality control. Readers looking for new methods or rigorous comparisons will not find much.\n\nI would not bring this to a reading group. I would not cite it. It does not deserve peer review because the evidence for the main performance claims is too thin to assess.","headline":"This is a routine fine-tuning of YOLOv8 plus standard TensorRT/OpenVINO acceleration on known defect datasets, with headline numbers that lack the controls needed to verify them.","tokens_in":2371,"tokens_out":351,"would_cite":false,"duration_ms":21682,"reading_group":"no","serious_thinker":"yes","would_accept_peer_review":false},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.3","headline":"Industrial-YOLO, a fine-tuned YOLOv8, exceeds 120 FPS on Jetson Orin hardware at 98.5 percent mAP on industrial defect datasets.","keywords":["defect detection","YOLOv8","edge computing","industrial inspection","real-time detection","NEU dataset","MVTec AD","TensorRT"],"falsifier":"Running the reported model and optimizations on the Jetson Orin with the NEU and MVTec test sets yields either under 120 FPS or under 98.5 percent mAP.","tokens_in":2619,"feed_emoji":"⚙️","tokens_out":652,"duration_ms":13939,"temperature":0.7,"pith_summary":"The paper introduces Industrial-YOLO as an edge-optimized framework based on fine-tuned YOLOv8 for detecting surface defects in manufacturing. It benchmarks the system on the NEU steel defect database, the MVTec AD dataset, and custom extensions for automotive and battery manufacturing that include scratches, pits, and inclusions. Target-specific optimizations using TensorRT and OpenVINO are applied to meet edge hardware constraints. The central demonstration is that these steps produce inference speeds above 120 FPS on the NVIDIA Jetson Orin while holding mean average precision at 98.5 percent. This setup is shown to support direct deployment on active assembly lines for automated optical inspection.","feed_headline":"Fine-tuned YOLOv8 exceeds 120 FPS on Jetson Orin for defect detection","feed_subtitle":"Keeps 98.5 percent mAP on NEU steel and MVTec datasets plus automotive extensions after TensorRT and OpenVINO optimizations.","key_machinery":"Fine-tuned YOLOv8 architecture accelerated by TensorRT and OpenVINO for edge deployment on NVIDIA Jetson Orin.","core_discovery":"Industrial-YOLO, built on a fine-tuned YOLOv8 architecture with TensorRT and OpenVINO optimizations, achieves inference speeds exceeding 120 FPS on the NVIDIA Jetson Orin platform while maintaining a mean Average Precision of 98.5 percent across the NEU surface defect database, MVTec AD, and custom automotive manufacturing extensions.","pith_inferences":["Similar edge optimizations could be applied to other YOLO variants or detection architectures on comparable hardware.","Testing on additional manufacturing lines beyond automotive would reveal how dataset-specific the accuracy holds.","Further hardware generations might push speeds higher while preserving the same accuracy level."],"forward_implications":["The framework supports zero-latency defect detection directly on active automotive assembly lines.","It provides a scalable approach for next-generation automated optical inspection systems in manufacturing.","High-speed performance holds across steel sheet defects and structural anomalies such as scratches, pits, and inclusions.","The same optimization pipeline can extend the approach to battery manufacturing datasets."],"fun_headline_variants":["YOLOv8 at 120 FPS on Jetson Orin with 98.5% mAP","YOLOv8 edge inference at 120 FPS with 98.5% mAP on Orin","Fine-tuned YOLOv8 at 120 FPS 98.5% mAP on Jetson Orin","YOLOv8 defect detection at 120 FPS on Jetson Orin 98.5% mAP"],"cache_read_input_tokens":2112,"weakest_assumption_plain":"The fine-tuning process and TensorRT/OpenVINO optimizations preserve the reported accuracy on the target industrial datasets without post-hoc data selection or hardware-specific retraining.","fun_headline_variants_meta":{"raw":{"variants":["YOLOv8 at 120 FPS on Jetson Orin with 98.5% mAP","YOLOv8 edge inference at 120 FPS with 98.5% mAP on Orin","Fine-tuned YOLOv8 at 120 FPS 98.5% mAP on Jetson Orin","YOLOv8 defect detection at 120 FPS on Jetson Orin 98.5% mAP"]},"model":"grok-4.3","cost_usd":0.01133,"raw_usage":{"total_tokens":4897,"prompt_tokens":676,"num_sources_used":0,"completion_tokens":107,"cost_in_usd_ticks":113303000,"prompt_tokens_details":{"text_tokens":676,"audio_tokens":0,"image_tokens":0,"cached_tokens":64},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":4114,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":676,"tokens_out":107,"duration_ms":43171,"temperature":1.0,"reasoning_tokens":4114,"cache_read_input_tokens":64,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-06-28T06:27:09.214706+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"Running the reported model and optimizations on the Jetson Orin with the NEU and MVTec test sets yields either under 120 FPS or under 98.5 percent mAP.","supporting_citations":[],"review_version":1}