{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:D7DWUI5JPMRCTJEMIVPUHTLNJO","short_pith_number":"pith:D7DWUI5J","schema_version":"1.0","canonical_sha256":"1fc76a23a97b2229a48c455f43cd6d4b92595ab491a0cf829a7890fa054d45f9","source":{"kind":"arxiv","id":"2605.14609","version":1},"attestation_state":"computed","paper":{"title":"Deep Image Segmentation via Discriminant Feature Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"A new loss based on classical discriminant analysis sharpens segmentation boundaries by making features more separable.","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Adam Dawid Sztamborski, Antonio Agudo, Ra\\\"ul P\\'erez-Gonzalo","submitted_at":"2026-05-14T09:25:36Z","abstract_excerpt":"Accurate image segmentation remains challenging, particularly in generating sharp, confident boundaries. While modern architectures have advanced the field, many of them still rely on standard loss functions like Cross-Entropy and Dice, which often neglect the discriminative structure of learned features, leading to inaccurate boundaries. This work introduces Deep Discriminant Analysis (DDA), a differentiable, architecture-agnostic loss function that embeds classical discriminant principles for network training. DDA explicitly maximizes between-class variance while minimizing within-class one,"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":true,"formal_links_present":true},"canonical_record":{"source":{"id":"2605.14609","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-05-14T09:25:36Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"8b12fce1a60d43dc963b15cbf7c64ef90026d74fb1ce16892e72c914f4deb8e7","abstract_canon_sha256":"fca97bee5d8b4c4cc85c25f8b881ccf926981f032b3e70933097ac5b901b7193"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:39:04.180826Z","signature_b64":"l/7b1QDbnmXRe4c55s1tgdPSkasPE6Q8p14qEIw5tpfehgQOZ1Fdyb/uOXPn+EVYypKK14jQYDySK8hMpOIKBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1fc76a23a97b2229a48c455f43cd6d4b92595ab491a0cf829a7890fa054d45f9","last_reissued_at":"2026-05-17T23:39:04.180103Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:39:04.180103Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Deep Image Segmentation via Discriminant Feature Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"A new loss based on classical discriminant analysis sharpens segmentation boundaries by making features more separable.","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Adam Dawid Sztamborski, Antonio Agudo, Ra\\\"ul P\\'erez-Gonzalo","submitted_at":"2026-05-14T09:25:36Z","abstract_excerpt":"Accurate image segmentation remains challenging, particularly in generating sharp, confident boundaries. While modern architectures have advanced the field, many of them still rely on standard loss functions like Cross-Entropy and Dice, which often neglect the discriminative structure of learned features, leading to inaccurate boundaries. This work introduces Deep Discriminant Analysis (DDA), a differentiable, architecture-agnostic loss function that embeds classical discriminant principles for network training. DDA explicitly maximizes between-class variance while minimizing within-class one,"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"DDA explicitly maximizes between-class variance while minimizing within-class one, promoting compact and separable feature distributions without increasing inference cost.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the observed improvements on the DIS5K benchmark are caused by the discriminant properties of the loss rather than by differences in training schedule, hyper-parameters, or implementation details not described in the abstract.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Deep Discriminant Analysis (DDA) is a new loss that maximizes between-class variance and minimizes within-class variance to produce more compact and separable features for image segmentation.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"A new loss based on classical discriminant analysis sharpens segmentation boundaries by making features more separable.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"e23f9a70586ba0cce06424f4016eb1e8fa3420bdb088fe922db95270f44f316b"},"source":{"id":"2605.14609","kind":"arxiv","version":1},"verdict":{"id":"b4f2888d-16ba-4c34-a88b-acbe4286d426","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T06:02:18.083261Z","strongest_claim":"DDA explicitly maximizes between-class variance while minimizing within-class one, promoting compact and separable feature distributions without increasing inference cost.","one_line_summary":"Deep Discriminant Analysis (DDA) is a new loss that maximizes between-class variance and minimizes within-class variance to produce more compact and separable features for image segmentation.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the observed improvements on the DIS5K benchmark are caused by the discriminant properties of the loss rather than by differences in training schedule, hyper-parameters, or implementation details not described in the abstract.","pith_extraction_headline":"A new loss based on classical discriminant analysis sharpens segmentation boundaries by making features more separable."},"references":{"count":38,"sample":[{"doi":"10.13039/501100011033","year":2026,"title":"Deep Image Segmentation via Discriminant Feature Learning","work_id":"65b3d757-cd63-4d82-92c8-70be0f6a295e","ref_index":1,"cited_arxiv_id":"2605.14609","is_internal_anchor":true},{"doi":"","year":null,"title":"It can lead to over- lapping foreground and background activations and uncertain or blurred boundaries","work_id":"c27c832d-9f33-47df-bc3b-285b2ad30f82","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"A largerS B indicates better class separability","work_id":"eda37088-e888-43b2-ba58-d68c313ba534","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Smaller values indicate more compact, discriminable classes","work_id":"bd527b60-15e7-44ef-ad78-03ec8e61147f","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"This desirable general separability criterion should take larger values when the within-class scatter is smaller and when the between-class scatter is larger","work_id":"fa2da256-ad58-4b90-9999-68cdff036769","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":38,"snapshot_sha256":"03b081cec002f78189b73181451db3eb3788471a303e3e6c9743280d2aa672f3","internal_anchors":2},"formal_canon":{"evidence_count":2,"snapshot_sha256":"1ba84952438072f731be2d919b5cc4668817913fde996c6498e1507925035a09"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2605.14609","created_at":"2026-05-17T23:39:04.180242+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.14609v1","created_at":"2026-05-17T23:39:04.180242+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.14609","created_at":"2026-05-17T23:39:04.180242+00:00"},{"alias_kind":"pith_short_12","alias_value":"D7DWUI5JPMRC","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_16","alias_value":"D7DWUI5JPMRCTJEM","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_8","alias_value":"D7DWUI5J","created_at":"2026-05-18T12:33:37.589309+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2605.14609","citing_title":"Deep Image Segmentation via Discriminant Feature Learning","ref_index":1,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":2,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/D7DWUI5JPMRCTJEMIVPUHTLNJO","json":"https://pith.science/pith/D7DWUI5JPMRCTJEMIVPUHTLNJO.json","graph_json":"https://pith.science/api/pith-number/D7DWUI5JPMRCTJEMIVPUHTLNJO/graph.json","events_json":"https://pith.science/api/pith-number/D7DWUI5JPMRCTJEMIVPUHTLNJO/events.json","paper":"https://pith.science/paper/D7DWUI5J"},"agent_actions":{"view_html":"https://pith.science/pith/D7DWUI5JPMRCTJEMIVPUHTLNJO","download_json":"https://pith.science/pith/D7DWUI5JPMRCTJEMIVPUHTLNJO.json","view_paper":"https://pith.science/paper/D7DWUI5J","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.14609&json=true","fetch_graph":"https://pith.science/api/pith-number/D7DWUI5JPMRCTJEMIVPUHTLNJO/graph.json","fetch_events":"https://pith.science/api/pith-number/D7DWUI5JPMRCTJEMIVPUHTLNJO/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/D7DWUI5JPMRCTJEMIVPUHTLNJO/action/timestamp_anchor","attest_storage":"https://pith.science/pith/D7DWUI5JPMRCTJEMIVPUHTLNJO/action/storage_attestation","attest_author":"https://pith.science/pith/D7DWUI5JPMRCTJEMIVPUHTLNJO/action/author_attestation","sign_citation":"https://pith.science/pith/D7DWUI5JPMRCTJEMIVPUHTLNJO/action/citation_signature","submit_replication":"https://pith.science/pith/D7DWUI5JPMRCTJEMIVPUHTLNJO/action/replication_record"}},"created_at":"2026-05-17T23:39:04.180242+00:00","updated_at":"2026-05-17T23:39:04.180242+00:00"}